Printer Friendly

Rules versus Home Rule Local Government Responses to Negative Revenue Shocks.

Local governments play an essential role in the provision of local public goods and deliver a wide range of government services. They are largely responsible for police protection and K-12 education, and they perform administrative functions such as providing building permits, issuing marriage licenses, and facilitating vehicle transfers. A failure to play these roles effectively can have dramatic consequences: a recent high-profile example is the water crisis in Flint, Michigan, where between 6,000 and 12,000 children were exposed to drinking water with high lead levels. This crisis was triggered by Flint's persistent financial dire straits, highlighting the challenges faced by local governments when dealing with negative revenue shocks. This paper studies both how cities respond to sudden negative shocks to revenue, and how the broader institutional framework shapes that response.

The increasing prevalence of e-commerce, which grew its share of all retail sales from 0.9 to 6.4 percent between 2000 and 2014 and continues to grow rapidly (Hortacsu and Sylverson, 2015), has made these questions more urgent. This rise erodes the tax base of large numbers of cities in the U.S., over half of which rely on local sales tax revenue (National League of Cities, 2014). Bruce et al. (2015) estimate that the loss in sales tax revenue due to the rise of e-commerce amounted to close to 4 percent of total sales tax revenue by 2012, and will continue to increase rapidly. Local sales taxes are a volatile source of revenue at business cycle frequencies as well.

Even property taxes, often thought to be a stable source of revenue for local governments, are susceptible to sizable shocks, as evidenced by the recent housing boom and bust (see Alm et al. (2014), Chernick et al. (2011), and Lutz et al. (2011)). In addition, the restrictions introduced by the Tax Cuts and Jobs Act on the deductability of state and local taxes will place downward pressure on revenue raised directly from individual residents. In order to effectively design local government policy in this context, it is crucial to understand how governments respond to negative revenue shocks, especially those that are likely to be permanent.

We carry out two empirical exercises in this paper. In the first one, we use national big-box chain bankruptcies that occurred during the Great Recession as natural experiments that allow us to analyze government responses to negative revenue shocks at the city level. We show that these bankruptcies provide a plausibly exogenous and discrete shock to local revenue, and we use that shock to study how expenditures respond. As individual big-box retailers typically account for roughly $20,000,000 in sales per year, a city losing one of these stores suffers a non-trivial hit just to sales tax revenue from the chain store in question alone. In addition, Shoag and Veuger (2018) show that after a big-box store shuts down, many other nearby businesses end up closing as well, exacerbating the consequences for local government finances. We compare cities that were home to the now defunct stores to cities where competitor retailers continued to operate to identify the causal impact of negative shocks to revenue on city budgets and behavior. We find that local governments that are hit by a big-box bankruptcy see their sales tax revenue decline by some 10 to 15 percent. In response, they decrease spending both on police protection and on administrative services.

We then proceed to investigate how the city's response ia moderated by its level of discretion, exploiting the fact that the degree of control that cities have over local policy varies both at the state level and within states. We study to what extent the size of the revenue drop varies with whether the city has "home rule" authority, an authority granted by state governments that allows some cities to implement certain policy changes without prior state approval, and we show that cities that are more constrained experience a sharper drop in revenue and slower rebounds in revenue than cities with more discretionary authority.

In our second empirical exercise, we exploit a feature of Illinois state law that automatically assigns home rule status to towns that surpass a population level of 25,000. We use a regression discontinuity design to show that the differences in the ability to respond to negative revenue shocks between cities with and without home rule are, at least in part, causal. Cities just above the cut-off endure less revenue volatility than their counterparts just below the cutoff, and enjoy stronger bond ratings.

The papers proceeds as follows. In section 1, we present our conceptual framework and hypotheses. We review the prior research on local government responses to negative shocks and how they are influenced by the level of autonomy enjoyed by local governments in section 2, where we also preview how we contribute to this existing body of research. We present our empirical settings and the data we use in section 3. The first empirical exercise is covered in section 4, while we turn to the regression discontinuity analysis in section 5. We conclude in section 6.

1 Conceptual Framework

Inspired by Tiebout's (1956) seminal article, much research on local public finance has focused on the provision of different bundles of local public goods. These local public goods needs to be paid for, and the typical starting point for analyses of their financing is that of Bradford and Oates (1971), that government revenue and private income are fungible. If the local government is hit by an unexpected shock to its revenue, in our case a negative one, voters will reoptimize. Assuming that there is no other change in the desirability of the various public and private goods, the logical response - and under certain political-economy assumptions, the predicted response - is to reoptimize and to raise new revenue to keep spending from falling as much as it would if it went down by the full amount of the negative revenue shock.

The amount of new revenue raised and the ways in which it is raised will be limited by three key factors. First, to the extent that the negative shock involves a negative shock to aggregate - public plus private - income, desired spending levels will now be lower than before. For a given amount of yearly revenue lost, this reduction will of course be larger if the shock is permanent. Assuming a marginal propensity to consume public goods of 5 cents per dollar of income, a permanent negative income shock of 1 dollar should permanently reduce desired revenue raised by 5 cents. We hypothesize that the type of large, negative shock to economic activity produced by a big-box bankruptcy will lower local-government revenue and spending overall. Subsection 4.1 shows that this is indeed the case.

Second, there is stickiness in the types of taxation and spending governments engage in. This phenomenon is in this context often referred to as the "flypaper effect," the idea that shocks have more of an effect where they hit (Gramlich, 1977; Fisher, 1982; Hines and Thaler, 1995). Subsection 4.2 demonstrates that the revenue losses observed in subsection 4.1 are indeed driven by reductions in sales and gross receipt tax revenue. This comes disproportionately from the retail industry, which is where the main blow landed. Localities respond to this reduction in revenue by cutting spending and raising revenue in other areas, in particular in areas where changes can be made relatively fast, such as cash holdings.

Third, there may be institutional and political constraints on how fast adjustments are made and whether they can be made at all. The institutional feature we focus on is whether a town enjoys "home rule." In the U.S. context, home rule is a term that refers to a greater level of autonomy local governments receive from their state. (1) Debates about whether local governments should have such greater autonomy usually touch on efficiency and effectiveness of different forms of governance. Home rule supporters argue that greater autonomy allows local citizens to address problems specific to their communities according to their preferences and with expedience (Tiebout, 1956). With home rule, local governments do not have to wait for approval from the state legislature or state officials to carry out policies. On the other hand, supporters of tighter state control over local governments argue that states can address local issues more effectively because they possess more technical expertise and can produce greater uniformity of governance and regulation (Richardson et al., 2003; Fajgelbaum et al., 2015).

This question of the appropriate level of decentralization is central to the literature on fiscal federalism (Musgrave (1959) and Oates (1972)). In general, the fiscal federalism literature argues that decentralized provision of public goods increases economic welfare by satisfying heterogeneous preferences across jurisdictions, albeit at the cost of a potential race to the bottom fueled by tax competition. Similarly, decentralized decision-making concerning fiscal policy generates greater efficiency in satisfying the varied circumstances unique to each municipality, but also introduces free-rider risk if local governments expect to be bailed out by higher levels of government (Oates, 1999). This combination of considerations leads us to predict that localities with home rule will recover faster from a negative revenue shock. We show, in subsection 4.3, that this is indeed the case: they manage to bring limit the reduction in own-source revenue after a negative shock relative to non-home rule cities. The impact of home rule on long-term fiscal health, on the other hand, is theoretically ambiguous. Our results in section 5 suggest that it is positive, that is, towns with home rule enjoy stronger bond ratings.

2 Literature Review

Two strands of the extensive literature that studies how governments respond to fiscal shocks are of particular relevance here: research on the flypaper effect, and on the institutional context. The evidence on the flypaper effect is decidedly mixed. For example, Gordon (2004) examines plausibly exogenous changes in Title I funding for school districts that occur shortly after the release of the Decennial Census. She finds evidence in support of the flypaper effect in the first year after the change: an increase in Title I funding leads to an increase in instructional spending. Three years out, however, localities adjust to the change in in-flows by decreasing revenue from other sources. This decrease in other revenue coupled with the increase in Title I funding yields a zero net change in instructional spending in the long-run. Knight (2002) shows that what looks like a flypaper effect in the context of the federal highway aid program is actually the result of grants being endogenous to spending priorities, while Lutz (2002) documents tax reductions that increase almost one for one with school grant receipts in New Hampshire.

On the other hand, Baicker (2004) finds that counties respond to sudden spending increases triggered by a capital crime conviction by contemporaneously raising taxes and cutting expenditures, specifically and in flypaper-type fashion, on public safety. Boylan and Ho (2017) find that the negative shock to state government finances induced by the Great Recession led to long-term cuts to education and health spending that were not undone during the recovery. These cuts did not simply eliminate wasteful spending but led to worse educational outcomes (Jackson et al., 2018) and are hard to fit into the Bradford-Oates framework. On the flipside, an example of a permanent positive revenue shock is studied by David and Ferreira (2017), who observe that rising housing prices between 1990 and 2009 caused a 20% increase in real per-pupil public-school spending. Singhal (2008) rationalizes flypapertype responses like these with a model of special-interest politics and confirms the existence of the phenomenon in the context of tobacco control policies. Leduc and Wilson (2017) provide additional evidence of such dynamics by showing that state-level highway spending increases in response to federal grants are greater in states with more political contributions from the public-works sector.

More directly related to the type of revenue shock we study, shocks to sales tax revenue that are likely to be permanent or at least different from business cycle fluctuations, is Agrawal's (2015) investigation of how local governments respond to the growing shift to e-commerce. Agrawal argues that because of enforcement problems and legal complications, the Internet serves as a sales tax haven. Using variation in Internet penetration, he finds that municipalities and states, chasing after disappearing revenue in a race to the bottom, reduced sales tax rates dramatically in response to the shift to online retail. These dynamics contributed to the rapid decline in sales tax revenue observed by Bruce et al. (2015). Diversification of revenue sources and use of rainy-day funds are examples of ways to deal with such unexpected negative shocks to a drop in a certain type of revenue.

We provide a comprehensive, nationwide set of estimates of the size and composition of cities' responses to such events. Initially, an unexpected sales tax loss translates to a fall in own-source revenue and spending cuts. Eventually, cities adjust by increasing revenue from property taxes, financial transactions, and charges or fees. This diversification response is strongly supported by the normative framework in Seegert (2016).

The broader institutional environment in which state and local governments operate has received its share of attention as well. For example, Poterba (1994) studies state responses to economic downturns and how those responses are influenced by state-level budget rules and politics. He finds that immediately after an unexpected budget deficit, states decrease spending. In subsequent years, states close the deficit with tax increases. This finding is echoed by Cromwell and Ihlanfeldt (2015), who find that Florida municipalities reacted to the loss of property tax and intergovernmental transfer revenue during the Great Recession by increasing property tax rates, and by reducing capital expenditures as well as non-essential public services. Follette and Lutz (2011) similarly provide evidence for pro-cyclical local government responses to downturns. Similar logic applies on the spending side and for positive revenue shocks.

Although there is a long tradition of economists using theory to weigh the pros and cons of home rule in taxation (e.g. Secrist, 1914), there is limited empirical research on how it shapes cities' fiscal policy. Most work in this area has come in the form of case studies at the city or state level, and the existing empirical evidence regarding the long-term fiscal health of cities with home rule is limited and mixed. Carroll and Johnson (2010), for example, find that towns in Connecticut and Maine, which have home rule, draw revenue from more diverse sources than towns in Minnesota and Vermont, which do not. Banovetz (2002) finds that 30 years of home rule in Illinois coincided with significant increases in tax rates as well as the adoption of new types of taxes, and that in some 5% of municipalities with home rule, voters, the courts, or the state legislature chose to retract that authority. That said, he argues that non-home rule municipalities, while not directly comparable, also witnessed tax hikes, and interprets the uncommon occurrence of repeal as support for home rule status. Similarly, Latzko (2008) notes that while Pennsylvania counties with home rule increased their spending more than non-home rules counties, he finds no evidence of higher tax rates in home rule counties.

We contribute to this literature in two ways. First, in section 4, we expand it to the national level and exploit plausibly exogenous local variation in shocks to local government revenue, from nationwide bankruptcies of big-box chains, to identify causal responses. We find that, as predicted, cities with home rule recover more rapidly from negative shocks to revenue by drawing from a broader range of revenue sources. The flypaper effect is thus muted in such towns, as we expected. Second, in section 5, we bring evidence to the table on the causal impact of home rule from variation in home rule assignment, as opposed to variation in which cities are hit by shocks. We do this by exploiting a discontinuity in Illinois law that makes it so that cities with a population over 25,000 are automatically given home rule. In this analysis, we find that home rule cities have better bond ratings and a greater degree of financial stability, which suggests that the benefits of flexibility outweigh the discipline imposed by rules issued by higher levels of government. Let us now turn to a discussion of these two empirical settings.

3 Empirical Setting

We will carry out two separate emprical exercises in the remainders of this paper. The first one features plausibly exogenous variation generated by bankruptcies of national big-box chains, while the second one relies on a regression discontinuity design that exploits a feature of the Illinois constitution. We assess the consequences of city home rule in both settings. In this section we will provide an introduction to that institutional feature first, before turning to aspects specific to each of our two testing grounds.

3.1 Home Rule

In the U.S. context, "home rule" is a term that generally refers to a greater level of autonomy local governments receive from their respective state through the state constitution, state legislation, or local charter (Richardson et al, 2003). As implied by this definition, local governments, such as counties, municipalities, and townships, derive their existence and power solely from their respective states, and home rule does not give them complete autonomy (Vanlandingham, 1968). This view is in line with the Tenth Amendment of the U.S. Constitution, which states that "[t]he powers not delegated to the United States by the Constitution, nor prohibited by it to the States, are reserved to the States respectively, or to the people." In addition, the U.S. Constitution does not contain any reference to local governments (ACIR, 1993). Therefore, only a state has the power to grant home rule to its local governments (ACIR, 1993), and each state's definition and implementation of home rule may differ from each other (National League of Cities, 2013).

What does this mean in practice? When questions regarding a local government's authority arise, the state constitution is consulted first, and if examining the state constitution does not resolve the issue, the courts will turn to the laws set by the state legislature. (2) If neither the state constitution nor the state law provides a clear answer, state courts decide. (3) There are two basic methods of interpretation: strict construction (also known as "Dillon's Rule") and liberal construction. Under strict construction, if a power is not enumerated among those granted to a local government, the local government does not have that power. On the other hand, under liberal construction, a local government possesses a certain power as long as it is not expressly taken from it. State constitutions and statutes can mandate either strict or liberal construction for different types of local governments.

These distinctions often date back quite some time. In the early republic, local governments enjoyed significant leeway in determining their own economic policies, which gave rise to "the patronage-based awarding of utility franchises; and (..) the deliberate creation and extinguishment of municipalities to avoid accumulated debt." This widespread corruption prompted Judge John F. Dillon of Iowa to formulate Dillon's Rule in Clark v. City of Des Moines (1865) and in City of Clinton v. Cedar Rapids and Missouri River Railroad (1868). Dillon's Rule established the guiding principle of strictly interpreting the scope of the local governments' power, and its growing popularity sparked debates over the level of autonomy local governments should have. Partially in response to these new strict construction practices, the home rule movement gained momentum in the late 19th century as states like Missouri (1875), California (1879), and Washington (1889) adopted constitutional home rule provisions that gave more autonomy to local governments. Over time, the back and forth over local autonomy has produced a range of combinations of devolved powers under home rule and of conditions under which localities qualify for home rule.

3.2 Bankruptcies of National Big-Box Chains

Our first empirical setting was born from corporate contretemps. In 2008, two major electronics retailers (Circuit City and CompUSA) and one major department store (Mervyn's) filed for bankruptcy and promptly liquidated the overwhelming majority of their existing stores. However, not all retailers in these categories failed. Best Buy, JC Penney, and Kohl's - a competing electronics big-box retailer and two major department store chains, respectively - continued to operate healthily. The chains that went bankrupt and the ones that continued to operate faced similar local business environments, and there is little to suggest that location choices (as opposed to prior corporate decisions) drove their fate, as discussed at length in Shoag and Veuger (2018). They show, among other things, that pre-trends in employment and business activity look similar for neighborhoods around eventually defunct and non-defunct stores; that bankruptcies had a large impact even if they control for zip-year effects, use variation in bankruptcy timing only, or allow for year-specific slopes for zip code level traits such as median house price; and that the neighborhoods look similar when we compare characteristics ranging from racial composition to access to public transit. All of this serves to sustain the idea that the negative revenue shocks induced by these big-box bankruptcies are orthogonal to local economic trends, that they are not the result of weak demand or slow population growth in the host cities, and that they are plausibly exogenous shocks to the localities' economies. Note that all of this is true even when we allow, among other things, for arbitrary trends within zip codes. This makes it exceedingly difficult to construct counterfactuals that can explain the patterns we observe in local business activity.

In addition, here, in Table 1, we test whether cities with stores from these two types of chains were on parallel trends in terms of different characteristics of localities' public finances. We use ESRI Business Analyst data supplied by Harvard University's Center for Geographical Analysis to calculate the number of Best Buy, Circuit City, CompUSA, JC Penney, Kohl's, or Mervyn's stores exist in each municipality in the U.S. in 2006. ESRI uses business data from InfoUSA, which compiles employment, sales, and location information on businesses in the United States, to construct its Business Analyst data. InfoUSA collects lists of establishments from phone directories, business filings, utility connections, press releases, web directories, annual reports, and other sources. It then surveys these establishments by phone (between 12 and 18 million establishments per year).

The financial characteristics tested include changes between 2005 and 2007 in total debt outstanding; in debt retired; in house prices; in property tax, charges and fees, and miscellaneous revenue; in own-source revenue; in state intergovernmental revenue; in total expenditures; and in sales tax revenue. To calculate these changes we use data from the U.S. Census of State and Local Government Finance for 2004 through 2012. The U.S. Census of State and Local Government Finance is conducted in full every five years (years ending in '2' and '7'). In other years, data is collected from a sample of local governments, and a new sample is chosen every five years (years ending in '4' and '9'). In all years, the Census collects data from in-sample local governments on revenues (taxes, charges, interest, etc.), total expenditures (education, health, public safety, infrastructure, etc.), debt, and financial assets. In our analysis, we include cities (by which we mean municipalities and townships) that are in the Census of State and Local Government Finances for at least one year pre-2008 and at least one year post-2008, and present in the data for at least five years. (4) In addition, we remove a municipality if the change between the minimum and maximum values of sales or total revenue, or of total expenditures, is greater than 500% to ensure that our results are not driven by outliers in terms of growth or by cities that fundamentally changed their tax system. We also drop one city with a population that appears to be miscoded, outlier cities with more than 50 of the big-box stores in our study (Houston, San Antonio, New York, and Los Angeles) (5), cities with zero of such big-box stores (as there can be no bankruptcies there), cities with more sales tax revenue than total revenue, and cities that had zero sales tax revenue during 2004-2007 (as there can be no negative sales tax revenue shock there). This leaves us with a sample of between 322 and 450 cities, depending on data availability for each variable.

Table 2 provides pre-2008 summary statistics for cities that were hit with bankruptcies and cities that were not. Along almost all dimensions, their pre-2008 finances are comparable in terms of per capita levels. Furthermore, as Table 1 shows, the bankruptcy variables (i.e. [BankruptDummy.sub.i] and [BankruptCount.sub.i]) are not associated with different changes in financial characteristics before the bankruptcies occurred, which further supports the claim that the two types of chains were located in cities that were on parallel paths in the years before 2008. One remaining concern is that spillovers, business activity displaced from a city with a bankruptcy to a nearby locality, could threaten our citylevel results. This turns out not to be a concern in practice, as our results are robust to the exclusion, as controls, of towns that are nearby or in the same county. Given all this, national-level bankruptcies allow us to identify the effects of negative revenue shocks on local government finances by deploying a difference-in-difference design, which we will do in section 4.

In subsection 4.3, we explore how these effects vary by home rule status. For nationwide data on this institutional feature, we draw from an ICMA (1974) survey, the U.S. Advisory Commission on Intergovernmental Relations (ACIR) (1993), and Krane et al. (2002). We use these sources to construct four distinct measures of home rule: two at the state level, and two at the city level. First, ACIR (1993) reports whether a state has granted structural home rule authority and/or broad functional home rule authority to the cities in that state. Cities with structural home rule authority are given the power to choose their own form of government, while those with functional home rule authority are given autonomy over local government functions such as taxation. According to ACIR, by 1993, forty states had granted their cities structural home rule authority, while only twenty-eight had granted them functional home rule authority. In our analysis, we use the functional home rule measure, as this type of home rule grants autonomy that is important to the type of decision making we focus on. Second (and third), Krane et al. (2002) also include information that indicates which states have structural, functional, and limited functional home rule. They report that thirty-one states had granted functional or limited functional home rule authority as of 2002. We use their data to construct a third measure as well, this one city-specific. Krane et al. (2002) detail the population each state requires a city to reach before it can be granted home rule status. Using these population limits, we can exclude nonhome rule cities that are in home rule states but that have not met the requirements for home rule authority. Finally, the ICMA (1974) survey gives us a city-level answer to the question "Within what type of charter or basic law does your city operate?," where the options were "unique charter," "uniform charter," "classification charter," "optional charter," "home rule," and "other." (6) We will present results based on the ACIR (1993) measure in the main text, and present robustness checks using the other three home rule concepts in appendix tables.

3.3 Home Rule in the Illinois State Constitution

Our second empirical setting is the state of Illinois. Illinois' state constitution states in article VII, section 6, that any municipality with a population above 25,000 is automatically given home rule authority. Municipalities with populations under this population cutoff can still elect via referendum to become home rule municipalities. Between 1970 and 2000 there were 191 referenda in Illinois, of which 97 passed and 94 failed (some of these latter towns also passed the 25,000 threshold during the same period). Note that towns generally do not lose home rule when their population decreases, and the existence of towns with home rule below the threshold is therefore not necessarily the result of strategic manipulation a la Eggers (2015). Conversely, a municipality with a population above the cutoff can, by referendum, elect to remove its home rule authority. Even though the population rule does not strictly determine home rule status, Figure 1 demonstrates that the probability a municipality has home rule does jump dramatically at the population cutoff of 25,000. In section 5, we exploit this break in home rule status in a fuzzy regression discontinuity design.

Illinois Comptroller's financial databases provide data on home rule status, population, and revenues of municipalities in Illinois. Appendix Figures 1 and 2 display the density of municipalities with a range of populations from 10,000 to 40,000 centered at the home rule population threshold in Illinois. The figures do not show a statistical break in the density of municipalities near the cutoff, which is evidence against endogenous sorting or manipulation of the running variable (McCrary, 2008). (7) With those considerations in mind, we use whether or not a city is above the population cutoff as an instrument for whether or not the city has home rule. The Illinois Comptroller database contains reports from 1994-2015. While we focus our analysis on recent years (2010-2015), we use the maximum population from 1994-2009 as our measure of population, as the maximum population is the relevant population for home rule determination. We ignore population post-2010 since it is endogenous to revenue changes.

Bond ratings for years 1994-1996 at the municipality-level are from the Illinois Comptroller's financial databases, and the more recent bond ratings were obtained by scrapping information from MunicipalBonds.com. In the data we use, 653 Illinois cities issue bonds (approximately half of the cities in the state). The percentage is even greater for cities from 5,000 to 45,000 in population: 80% of the sample of cities in that range, or 266 total. Bond ratings are not available for all of these cities. About 240 cities in the bond data do not have a bond rating available and 67 cities in our sample bandwidth do not have a bond rating available. We code these cities as not having an extremely strong bond rating. This decision does not affect the results. First, the probability that a city is missing a bond rating does not jump discretely at the 25,000 home rule threshold. Second, if we instead code those cities as missing for the bond analysis, the results are qualitatively similar to the results reported below in section 5.

4 Revenue Shocks from Big-Box Bankruptcies

In this first empirical exercise, we compare cities that were home to a bankrupt chain to cities that were home to a surviving chain by analyzing their finances before and after the bankruptcies in 2008. We will look first at the size and persistence of the revenue shocks, then at how cities' finances responded to these shocks, and finally at how this response varied by home rule states.

4.1 Size and Persistence of the Revenue Shocks

To see what happened to city revenue after the big-box bankruptcies, we run regressions of the following kind:

ln([Revenue.sub.it.sup.h]) = [alpha] + [[beta](BankruptDummy * Post).sub.it] + [theta][(OperatingDummy *Post).sub.it] +[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it] (1)

where [Revenue.sup.h.sub.it] is revenue in category h, where the category is either sales tax and gross receipt revenue or own-source revenue in municipality i in year t. Own-source revenue captures all revenue generated by the municipality itself and does not include intergovernmental revenues. [BankruptDummy.sub.i] equals 1 if city i contained either a Circuit City a CompUSA, or a Mervyn's, and 0 otherwise. [Post.sub.t] is an indicator for whether or not the year is after 2008, the bankruptcy year. The interaction term is our variable of interest.

We control for whether or not municipality i contains any operating stores after the bankruptcy year using the interaction term formed by [OperatingDummy.sub.i] and [Post.sub.t] (8). Finally, [[delta].sub.i] represents municipality fixed effects and [[gamma].sub.t] represents time fixed effects. Standard errors are clustered at the county level. Our unit of observation is the city-year. We have 539 cities and 9 years of data. The total sample is 4,350 city-year observations. Note that the sample is not 539x9 = 4,851 because our sample restriction only requires that cities have at least 5 years of data and be present pre- and post-2008. We also estimate similar regressions that contain counts of the number of bankrupt big-box stores instead of the dummy variable shown in equation 1. Note also that the identifying assumption here is that, conditional on the included covariates, fiscal outcomes would have evolved similarly across the two types of localities in the sample had the national-chain bankruptcies not occurred. As we saw, Table 1 suggests that this would indeed have been the case. (9)

Table 3 shows our estimates of the effect the bankruptcies had on local revenue. The first row of Panel A of Table 3 shows that municipalities suffered a loss of between 9% and 16% of local sales tax revenue and gross receipt revenue, depending on whether we include state by year fixed effects. While a single bankruptcy, even the bankruptcy of a big-box retailer, is unlikely to cause such a large decline, Shoag and Veuger (2018) show that significant numbers of stores located close to a Circuit City, CompUSA, or Mervyn's store shut down as a consequence of their disappearance. The second row, where BankruptCount (the sum of Circuit City, CompUSA, and Mervyn's stores in a city) replaces [BankruptDummy.sub.i], shows that for each big-box store going bankrupt, a municipality's sales tax and gross receipt revenue went down by 1.6% to 4.3%. The first row of Panel B from Table 3 shows that because the municipalities in our sample rely heavily on local sales tax revenue, this shock actually translates into a significant dent in own-source revenue, with decreases of between 4.0% and 5.0%. The second row paints a similar picture; for each big-box store going bankrupt, a municipality's own-source revenue will go down by about 1.8%. All of these results are robust, at least in terms of direction, order of magnitude, and statistical significance, to the inclusion of state-by-year fixed effects, which is of particular interest because it demonstrates that they are not driven by the differential impact of the Great Recession across the country. (10)

We test the persistence of the shocks to revenue by interacting the bankruptcy dummy variable with dummy variables for the year before the bankruptcy and the four years after. Panel A of Table 4 shows that the shocks to sales tax revenue (i.e. the reduction of sales tax and gross receipt revenue of about 7% to 12%) persisted even four years after the bankruptcy, perhaps because municipalities struggled to fill empty store fronts, or because customers switched to online shopping permanently. In fact, the effect of bankruptcy becomes 1% to 3% more severe from year 1 to year 4 after the bankruptcies. Interestingly, Panel B from Table 4 shows that the decline in own-source revenue decays within one or two years, as municipalities turn towards other sources of revenue for the loss. The next subsection sheds light on that development.

4.2 Local-Government Financial Response

We now turn our attention to the way in which local policymakers respond to the drops in revenue observed above. Let us first look at spending. We run regressions of the following form:

ln([Expenditure.sup.h.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it] +[[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it] (2)

where [Expenditure.sup.h.sub.it] is the amount of local government expenditures in category h, where the categories are total expenditures, police protection, capital outlays, financial administration, total debt outstanding, and cash securities. Panel A from Table 5 shows estimates of the drop in four of the six categories, with the most severe reductions in financial administration (about 10%) and cash securities (about 7%). The estimate for total expenditures, a 3.36% decrease, is only slightly smaller than the effect we found on own-source revenue (3.41% decrease). As to the type of expenditures that are cut, we confirm the findings of Baicker (2004) and Cromwell and Ihlandfeldt (2015): cities decrease spending on police protection and administrative services. Panel B of Table 5, which replaces Bankrupt Dummy with Bankrupt Count, shows similar results; the more big-box stores went bankrupt, the higher the reduction on various expenditures. (11)

Turning back to the revenue side, Figure 2a shows that there is a statistically significant reduction in sales tax and gross receipt revenue and own-source revenue generally immediately after the bankruptcy year, as implied by Table 3. The pre-trends suggest that this is a causal consequence of the bankruptcies. However, when examining the more specific components of municipalities' own-source revenues, Panel C of Table 5 shows that big-box shocks actually have a positive effect on financial transactions or property tax combined with financial transactions. We see here how municipalities immediately attempt to turn to other sources of revenue as their sales and gross receipt tax revenue declines. This result partially explains why the effect of the bankruptcies on own-source revenues is not as negative as the effect on sales tax and gross receipt revenue. This difference in magnitude is also partially mechanical: sales tax revenue is only a fraction of own-source revenue, thus any decrease in sales tax revenue should lead to a smaller and proportional decrease in own-source revenue. In Panel D, we show estimates of the effect of bankrupt count on these revenues. The coefficients are positive but smaller in magnitude and not statistically distinguishable from zero. We now turn to an analysis of how the broader institutional environment affects these responses.

4.3 Home Rules Status and Local Government Responses

To explore the consequences of these differences in policy instrument availability, we run regressions of the following type:

ln([Revenue.sup.h.sub.it]) = [alpha] + [[??](BankruptDummy * Post * HomeRule).sub.it] + [[beta](BankruptDummy * Post).sub.it] +[[rho](HomeRule * Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it] (3)

where [Revenue.sup.h.sub.it] is total local government revenue in category h, where h is either sales tax and gross receipts revenue or own-source revenue. [BankruptDummy.sub.i] is a dummy variable equal to 1 if municipality i has a big-box store that will go bankrupt (i.e. Circuit City CompUSA, or Mervyns) and equal to 0 if municipality i does not have a big-box store that will go bankrupt but has one of the comparison stores (i.e. Khols, JC Penney, or Best Buy), [Post.sub.t] is an indicator for whether or not year t is after 2008 (the bankruptcy year), while [HomeRule.sub.i] is a dummy variable equal to 1 if the city has home rule status based on ACIR (1993) (12). The triple interaction is our variable of interest. In addition, we control for all other interactions either directly or through municipalities fixed effects. We also control the still-operating stores a city has after 2008 and we include year fixed effects.

Column 1 of Table 6 shows no robust, statistically significant difference in sales tax and gross receipt revenue between municipalities that enjoy home rule and municipalities that do not, which is unsurprising: there is no reason why policy flexibility should shield you from the kind of negative revenue shock that a big-box store bankruptcy triggers. For ease of comparison, we include a row labeled "Combined Effect" that represents the total effect of the bankruptcy ([??] + [beta]) in home rule municipalities. This can be compared to the row labeled "Bankrupt Dummy" or "Bankrupt Count" that represents the total effect of the bankruptcy ([beta]) in municipalities without home rule. Panel B of Table 6, which presents the results for robustness tests that replace Bankrupt Dummy with Bankrupt Count, supports the same conclusion.

However, there is reason to believe that home rule allows you to recover more swiftly, and we see evidence of that here. Cities with home rule status face smaller declines in own-source revenue after the shock. In fact, column 2 in Panel A of Table 6 shows municipalities with home rule status experience a reduction in own-source revenue that is 73% less severe than that experienced by those without home rule when we rely on our bankruptcy dummy estimator. Comparing column 2 in Panels A and B shows that on the intensive margin - when we take into account the number of bankruptcies - the impact of home rule is similar.

Column 3 of Table 6 shows the mechanism through which this happens, at least partially: through property taxes, financial-market revenue, and miscellaneous revenue. Out of all the municipalities experiencing bankruptcies, the ones with home rule are able to raise about 13% more property tax, charges and fees, and miscellaneous revenue than the ones without. Figure 2b shows this dynamically, as revenue in home rule cities recovers more quickly than in cities without such flexibility. Column 4 of Table 6 shows that home rule and non-home rule cities are not statistically different in terms of post-bankruptcy total expenditures. However, the estimated effect of bankruptcy on spending in home rule cities is approximately 20% to 30% of the estimated effect of bankruptcy on spending in non-home rule cities. The magnitude of this difference is similar to the difference in own-source revenue loss for home rule versus non-home rule cities.

Now, one may worry that home rule amendments are common to states in a particular region of the country, and that cities in that region rapidly recovered from the big-box store bankruptcies for other reasons. We estimate a series of regressions where we interact regional dummies with the bankruptcy count variable. Results are presented in Table A.13. Column 1 replicates column 2 from Panel A of Table 6 using the Krane et al. (2002) home rule definition. Column 2 controls for (interactions with) Census region, column 3 for Census division, while column 4 controls for state. The coefficients remain statistically significant and around 10% to 15%, except for the coefficient from column 3 that goes down to about 6.2% and is no longer significant at the 1% level. Overall, the results are similar and our conclusions remain unchanged. To strengthen our case that they are unlikely to be driven by unobserved, systematically different features of cities with and without home rule, we turn to the second leg of our empirical analysis, a regression discontinuity analysis of cities in Illinois.

5 Regression Discontinuity Analysis of Home Rule Status and Revenue Stability

While our first empirical exercise relied on quasi-random assignment of negative shocks to cities, our second exercise relies on quasi-random assignment of home rule status in Illinois, as discussed in section 3. This exercise serves to address concerns that cities with home rule are systematically different not in the shocks that they are hit with, but in how they respond to such shocks beyond differences generated by their home rule status. Our preferred specification is a local linear fuzzy regression discontinuity design with a triangular kernel, which places the most weight on those cities closest to the population cutoff of 25,000 above which cities acquire home rule. We estimate the model using a range of population bandwidths (from +\- 12,500 to +\- 20,000), and in all cases, the results are qualitatively the same. (13) Specifically, we estimate the following first-stage regression on the sample of municipalities near the discontinuity:

[HomeRule.sub.i] = [alpha] + [[beta](Above25000).sub.i] + [[theta](Population).sub.i] + [[rho](Above25000 * Population).sub.i] + [[epsilon].sub.i] (4)

Results are shown in Panel A of Table 7. We see that municipalities with a population over 25,000 are about 60% more likely to have home rule authority.

The second stage produces an estimate of the causal effect home rule status has on revenue stability, as follows:

[RevFall.sub.i] = [alpha] + [[beta](HomeRule).sub.i] + [[rho](Population).sub.i] + [[lambda] (HomeRule * Population).sub.i] + [[epsilon].sub.i] (5)

where [RevFall.sub.i] is the largest annual percentage fall in revenue from 2010 to 2015 in municipality i. Note that [RevFall.sub.i] is strongly correlated with the standard deviation in per capita revenue from 2010-2015; the correlation is about 0.5 after excluding outliers above the 99th percentile in both variables. In other words, places with bigger revenue falls also have more variation in per-capita revenue over the full time period. We believe that [RevFall.sub.i] is a better measure of volatility since we argue and find that home rule municipalities are good at forestalling and curtailing revenue shocks. Case in point: Panel B shows that home rule makes revenue reduction 8% to 10% less severe. Panel C replaces [RevFall.sub.i] with [RevFall10.sub.i], which is a dummy variable that equals 1 if municipality i has a fall in revenue larger than 10% at any point from 2010 to 2015. Municipalities with home rule are about 20% to 30% less likely to experience revenue reduction greater than 10%. Panel D replaces [RevFall.sub.i] with [RevFall30.sub.i], which is a dummy variable that equals 1 if municipality i had a fall in revenue larger than 30% at any point from 2010 to 2015. Again, municipalities with home rule status are significantly less likely (18% to 19% less likely) to experience a dramatic fall in revenue. Taken together, these results suggest that home rule municipalities are not as vulnerable to sharp revenue downturns as non-home rule municipalities.

Last but not least, Panel E replaces [RevFall.sub.i] with [StrongBond.sub.i], which is a dummy variable that equals 1 if municipality i has an extremely strong (triple A) bond rating. The regression results in Panel E show that municipalities with home rule are about 35% more likely to have an extremely strong bond rating. (14) This inference is, of course, only valid locally, but over 40% of the population in Illinois lives in municipalities with populations between 5,000 and 45,000 (i.e. populations within 20,000 of the threshold). Excluding Chicago, over 60% of people in Illinois live in those municipalities. With nearly half of the state's population (over 5 million people as of 2010) living in municipalities close to the home rule threshold, we believe that even the local inference is important and relevant for policy. (15)

6 Discussion

As e-commerce becomes more dominant, local governments are likely to continue experiencing revenue shifts similar to those produced by the bankruptcies of big-box retail chains during the Great Recession. The negative shocks generated by bankruptcies during this period of aggregate-demand shortfalls were most likely greater and, in particular, more durable than they would have been during normal times. While this may make them less directly comparable to business cycle driven shocks during other periods, it makes them more informative as we think through the effects of shocks induced by long-term structural transformations in the economy. The results above offer robust estimates of the effect of such shocks on revenue and expenditures, of local governments' responses, and of the importance of the legal framework cities operate in.

In addition, we demonstrate that municipalities with less discretionary decision-making, i.e. no home rule, experience a sharper drop in revenue and a slower rebound in revenue than municipalities with more discretionary authority. The downside of autonomy typically considered in the literature on rules versus discretion is a lack of credibility and self-control. In the results of our regression discontinuity analysis of cities in Illinois we do not see evidence of home rule towns' bond ratings being worse, while our regression results suggest that their spending bounces back faster from negative shocks. This suggests to us that home rule cities are not more likely to live beyond their means: if anything, they are more fiscally responsible, suggesting that in this case, discretion trumps rules. This may be the case, in part, because of the constitutional restrictions placed by cities on their own spending and taxing abilities that Brooks et al. (2016) analyze. They find that these self-imposed home rule-type restrictions do indeed reduce municipal revenue growth.

These findings illustrate the upside of granting policymakers discretion, as opposed to tying their hands. In that sense, and despite the fact that this paper deals most directly with a question about local government responses to revenue shocks, we contribute to a larger literature on rules versus discretion started by Kydland and Prescott (1977). This is important in the context of both federalist systems like the United States and supranational organizations like the European Monetary Union. In that literature, with the upside of flexiblity comes potential downsides: a loss of focus on the long run and a loss of credibility. Similarly, while state governments recognize that home rule status can be a source of helpful flexibility in times of crisis, this is often coupled with concerns that giving local politicians too much leeway will result in financial distress in the long run. Our results suggest that one ought not worry about that too much.

References

ACIR (1993) State Laws Governing Local Government Structure and Administration. Washington, D.C.: U.S. Advisory Commission on Intergovernmental Relations, March.

Agrawal, David R. (2015) "The Internet as a Tax Haven? The Effect of the Internet on Tax Competition," Mimeo: University of Georgia.

Alm, James, Robert D. Buschman, and David L. Sjoquist (2014) "Foreclosures and Local Government Revenues from the Property Tax: The Case of Georgia School Districts," Regional Science and Urban Economics, 46: 1-11.

Baicker, Katherine (2004) "The Budgetary Repercussions of Capital Convictions," Advances in Economic Analysis and Policy, 4(1): 1-26.

Banovetz, James M. (2002) "Illinois Home Rule: A Case Study in Fiscal Responsibility," Journal of Regional Analysis and Policy, 32(1): 79-98.

Boylan, Richard, and Vivian Ho (2017) "The Most Unkindest Cut of All? State Spending on Health, Education, and Welfare During Recessions," National Tax Journal, 70(2): 329-66.

Bradford, David F., and Wallace E. Oates (1971) "The Analysis of Revenue Sharing in a New Approach to Collective Fiscal Decisions," Quarterly Journal of Economics, 85(3): 416-39.

Brooks, Leah, Yosh Halberstam and Justin Phillips (2016) "Spending Within Limits: Evidence from Municipal Fiscal Restraints," National Tax Journal, 69(2): 315-52.

Bruce, Donald, William F. Fox, and LeAnn Luna (2015) "E-tailer Sales Tax Nexus and State Tax Policies," National Tax Journal, 68(3S): 735-66.

Carroll, Deborah A., and Terri Johnson (2010) "Examining Small Town Revenues: To What Extent Are They Diversified?" Public Administration Review, 70: 223-35.

Chernick, Howard, Adam Langley, and Andrew Reschovsky (2011) "The Impact of the Great Recession and the Housing Crisis on the Financing of America's Largest Cities," Regional Science and Urban Economics, 41(4): 372-81.

Coester, Adam (2004) "Dillon's Rule or Not?" National Association of Counties, January.

Constitution of the State of Illinois. Art. VII, Sec. 6.

Cromwell, Erich W., and Keith R. Ihlanfeldt (2015) "Local Government Responses to Exogenous Shocks in Revenue Sources: Evidence from Florida," National Tax Journal, 68(2): 339-76.

Davis, Matthew, and Fernando V. Ferreira (2017) "Housing Disease and Public School Finances," NBER Working Paper No. 24140.

Eggers, Andrew C., Ronny Freier, Veronica Grembi, and Tommasso Nannicini (2015) "Regression Discontinuity Designs Based on Population Thresholds: Pitfalls and Solutions," IZA Discussion Paper Series No. 9553.

Fajgelbaum, Pablo D., Eduardo Morales, Juan Carlos Suarez Serrato, and Owen M. Zidar (2015) "State Taxes and Spatial Misallocation." NBER Working Paper No. 21760.

Fisher, Ronald (1982) "Income and Grant Effects on Local Expenditures: The Flypaper Effect and Other Difficulties," Journal of Urban Economics, 12(3): 324-45.

Follette, Glenn, and Byron F. Lutz (2011) "Fiscal Policy in the United States: Automatic Stabilizers, Discretionary Fiscal Policy Actions, and the Economy." In: Follette, Glenn, and Byron F. Lutz, Fiscal Policy: Lessons from the Crisis. Perugia: Banca d'Italia, February.

Gordon, Nora (2004) "Do federal grants boost school spending? Evidence from Title I," Journal of Public Economics, 88(9): 1771-92.

Gramlich, Edward M. (1977) "Intergovernmental Grants: A Review of the Empirical Literature," in W. E. Oates (ed.), The Political Economy of Federalism, Lexington, MA: Lexington Books, 219-40.

Hines, James R. Jr., and Richard H. Thaler (1995) "Anomalies: The Flypaper Effect," Journal of Economic Perspectives, 9(4): 217-26.

Hortacsu, Ali, and Chad Syverson (2015) "The Ongoing Evolution of US Retail: A Format Tug-of-War," Journal of Economic Perspectives, 29(4): 89-112.

ICMA (1974) Municipal Electoral Systems and City Council Structure. Washington, D.C.: International City/County Management Association.

Jackson, C. Kirabo, Cora Wigger, and Heyu Xiong (2018) "Do School Spending Cuts Matter? Evidence from the Great Recession," NBER Working Paper No. 24203.

Knight, Brian (2002) "Endogenous Federal Grants and Crowd-out of State Government Spending: Theory and Evidence from the Federal Highway Aid Program," American Economic Review, 92(1): 71-92.

Krane, Dale, Melvin B. Hill, and Platon N. Rigos (2002) Home Rule in America: A Fifty-State Handbook. Washington, D.C.: CQ Press.

Kydland, Finn E., and Edward C. Prescott (1977) "Rules Rather than Discretion: The Inconsistency of Optimal Plans," Journal of Political Economy, 85(3): 473-92.

Latzko, David A. (2008) "Home Rule and the Size of County Government in Pennsylvania," Journal of Regional Analysis and Policy, 38(1): 89-96.

Leduc, Sylvain, and Daniel Wilson (2017) "Are State Governments Roadblocks to Federal Stimulus? Evidence on the Flypaper Effect of Highway Grants in the 2009 Recovery Act," American Economic Journal: Economic Policy, 9(2): 253-92.

Lutz, Byron (2010) "Taxation with Representation: Intergovernmental Grants in a Plebiscite Democracy," Review of Economics and Statistics, 92(2): 316-32.

Lutz, Byron, Raven Molloy, and Hui Shan (2011) "The Housing Crisis and State and Local Government Tax Revenue: Five Channels," Regional Science and Urban Economics, 41(4): 306-19.

McCrary, Justin (2008) "Manipulation of the Running Variable in the Regression Discontinuity Design: A Density Test," Journal of Econometrics, 142(2): 698-714.

Musgrave, Richard A. (1959) The Theory of Public Finance. McGraw-Hill Book Company, Inc.

National League of Cities (2013) Local Government Authority. Retrieved from http://www.nlc.org/build-skills-and-networks/resources/cities-101/city-powers/local-government-authority.

National League of Cities (2014) City Fiscal Conditions in 2014. National League of Cities, Center for City Solutions and Applied Research.

Oates, Wallace E. (1972) Fiscal Federalism. Harcourt Brace Jovanovich.

--(1999) "An Essay on Fiscal Federalism," Journal of Economic Literature, 37(3): 1120-49.

Poterba, James M. (1994) "State Responses to Fiscal Crises: The Effects of Budgetary Institutions and Politics," Journal of Political Economy, 102(4): 799-821.

Richardson, J. J., M. Z. Gough, and R. Puentes (2003) "Is Home Rule the Answer?: Clarifying the Influence of Dillon's Rule on Growth Management," Center on Urban and Metropolitan Policy, the Brookings Institution.

Roback, Jennifer (1982) "Wages, Rents, and the Quality of Life," Journal of Political Economy, 90(6): 1257-78.

Rosen, Sherwin (1979) "Wages-based Indexes of Urban Quality of Llife," in P. Mieszkowski and M. Straszheim (Eds.), Current Issues in Urban Economics. Baltimore, MD: John Hopkins University Press.

Secrist, Horace (1914) "Home Rule in Taxation," Quarterly Journal of Economics, 28(3): 490-505.

Seegert, Nathan (2016) "The Causes and Consequences of Increasing State Tax Revenue Volatility," Mimeo: University of Utah.

Shoag, Daniel W., and Stan A. Veuger (2018) "Shops in the City: Evidence on Local Externalities and Local Government Policy from Big-Box Bankruptcies," Review of Economics and Statistics, 100(3): 440-453.

Singhal, Monica (2008) "Special Interest Groups and the Allocation of Public Funds," Journal of Public Economics, 92(3-4): 548-64.

Tiebout, Charles M. (1956) "A Pure Theory of Local Expenditure," Journal of Political Economy, 64(5): 416-24.

Vanlandingham, Kenneth E. (1968) "Municipal Home Rule in the United States," William & Mary Law Review, 10, Rev. 269.

Daniel Shoag ([dagger]), Cody Tuttle ([double dagger]), and Stan Veuger ([section])

(*) We thank Nikolai Boboshko, Nick Carollo, Philip Hoxie, and Hao-Kai Pai for excellent research assistance. Stacy Dickert-Conlin, Bill Gentry, Mark Skidmore, and three anonymous referees made tremendously helpful suggestions for improvement. Seminar attendees at the Annual Conference of the Southern Economic Association, the Cato Institute, the Greater Boston Area Urban and Real Estate Economics Seminar, the National Tax Association's Annual Conference on Taxation, the Spatial Economics Research Annual Conference, and the University of Maryland, College Park, provided important comments as well. We thank the Laura and John Arnold Foundation for generous financial support.

([dagger]) Weatherhead School of Management, Case Western Reserve University and Harvard Kennedy School, Harvard University

([double dagger]) University of Maryland, College Park

([section]) Corresponding Author: 1789 Massachusetts Avenue, Washington, DC 20010, stan.veuger@aei.org; American Enterprise Institute for Public Policy Research, IE School of Global and Public Affairs, and Tilburg University

(1) We discuss home rule as a legal and institutional construct in more detail in section 3.

(2) This paragraph and the following three rely heavily on Richardson et al. (2003), and quotations originate there.

(3) A court's interpretation can of course be overruled if an amendment to the state constitution or if a new law enacted by the state legislature provides clear instructions on how to solve the issue.

(4) Table A.1 shows that these sampling restrictions do not affect our results.

(5) We drop the four largest cities out of concern that they are large enough to drive even national bankruptcies. Obviously any cutoff is arbitrary, but our results do not vary much with our choice of cutoff.

(6) Table A.2 presents summary statistics for both home rule and non-home rule cities.

(7) Table A.3 shows summary statistics by population.

(8) Cities with more bankrupt stores maybe more likely to have continually operating stores. Since the presence of operating stores may have a time-varying effect on revenue, that effect will not be absorbed by city fixed effects, which is why we include this control variable in our preferred specification. The control ends up being mostly irrelevant from an empirical perspective: panel A of Appendix Table 4 shows that whether we include this control or not does not affect our results qualitatively.

(9) These results, and those in the rest of the paper that rely on the same approach, are not qualitatively different when we use a matching estimator that relies on Coarsened Exact Matching based on the municipalities' population and the number of big-box stores.

(10) In Table A.4, we demonstrate that our main results are robust to eliminating the OperatingDummy term, the addition of state-level controls, and the aggregation up to the county level.

(11) Tables A.5 through A.8 and Figure A.3 show that these spending cuts are not the result of population declines. Cities with a bankruptcy undergo only statistically insignificant declines in population that are almost an order of magnitude smaller than the spending cuts we observe, and including population controls does not materially affect our estimates. Figure A.4 shows event study graphs for all outcome variables discussed in this subsection; there does not appear to be a general pattern of pre-shock trend differentials that would pose a threat to our identification strategy.

(12) Robustness checks using our three alternative home rule measures can be found in Tables A.9 through A.12

(13) Table A.14 shows a number of additional bandwiths as robustness checks.

(14) Figures 3a, 3b, and 4 are graphical representations of the regression results from Panels B, C, and E, respectively, from Table 7.

(15) Tables A.14 and A.15 show that our estimates are reasonably robust to different specifications.
Table 1: Evidence of Parallel Pre-trends from 2005-2007

Panel A         (1)          (2)              (3)
                '05-'07      '05-'07          '05-'07
                Total Debt   Debt Retired     House Prices
                Outstanding

Bankrupt Dummy    0.0422       0.0321          -0.0163
                 (0.0361)     (0.0883)         (0.0141)
Observations    448          446              322
                 (5)          (6)              (7)
                '05-'07      '05-'07          '05-'07 Total
                Own-Source   State Intergov.  Expenditures
                Revenue      Revenue
Bankrupt Dummy    0.0235      -0.0477           0.0213
                 (0.0145)     (0.1710)         (0.0176)
Observations    450          442              450
Panel B          (1)          (2)              (3)
                '05-'07      '05-'07          '05-'07
                Total Debt   Debt Retired     House Prices
                Outstanding
Bankrupt Count    0.0005       0.0276          -0.0016
                 (0.0068)     (0.0242)         (0.0032)
Observations    448          446              322
                 (5)          (6)              (7)
                '05-'07      '05-'07          '05-'07 Total
                Own-Source   State Intergov.  Expenditures
                Revenue      Revenue
Bankrupt Count    0.0017       0.0180          -0.0030
                 (0.0031)     (0.0429)         (0.0052)
Observations    450          442              450

Panel A         (4)
                '05-'07
                Property Tax,
                Charges & Fees, &
                Misc. Rev.

Bankrupt Dummy    0.0266
                 (0.0231)
Observations    450
                 (8)
                '05-'07
                Sales Tax Revenue
Bankrupt Dummy   -0.0283
                 (0.0517)
Observations    423
Panel B          (4)
                '05-'07
                Property Tax,
                Charges & Fees, &
                Misc. Rev.
Bankrupt Count   -0.0025
                 (0.0057)
Observations    450
                 (8)
                '05-'07
                Sales Tax Revenue
Bankrupt Count   -0.0112
                 (0.0094)
Observations    423

Note: Panel A of this table reports estimates of regressions of the
following form:

(ln([X.sub.2007]) - ln([X.sub.2005])) = [alpha] +
[beta][BankruptDummy.sub.i] + [[epsilon].sub.i]

where [X.sub.t] is a financial characteristic of a municipality
measured in year t. Eight financial characteristics are included, which
are total debt outstanding; debt retired; house prices; property tax,
charges and fees, and miscellaneous revenue; own-source revenue; state
intergovernmental revenue; total expenditures; and sales tax revenue.
[BankruptDummy.sub.i] is equal to 1 if municipality i has a big-box
store that will go bankrupt (i.e. Circuit City, CompUSA, or Mervyns)
and equal to 0 if it does not have a big-box store that will go
bankrupt but does have one of the comparison stores (i.e. Kohls, JC
Penney, or Best Buy). Panel B reports estimates of regressions of the
following form:

(ln([X.sub.2007]) - ln([X.sub.2005])) = [alpha] + [beta]
[BankruptCount.sub.i] + [[epsilon].sub.i]

where [BankruptDummy.sub.i] is replaced with [BankruptCount.sub.i],
which is equal to the number of big-box stores that will go bankrupt in
municipality i and equal to zero if the municipality does not have a
big-box store that will go bankrupt but does have one of the comparison
stores. Standard errors are clustered at the county-level in all of the
regressions in this table, except for the regression with state
intergovernmental revenue (column 6) where the standard errors are
clustered at the state-level.

(*) p < 0.1, (**) p< 0.05, (***) p< 0.01

Table 2: Per Capita Differences Between Cities with Defunct and
Operational Chains, pre-2008

                              No     Bankrupt    Difference:
                           Bankrupt   Stores    (No Bankrupt
                            Stores               - Bankrupt)

Total Revenue                1.991     2.049     -0.058
                            (1.487)   (1.280)    (0.120)
Own-Source Revenue           1.714     1.738     -0.024
                            (1.192)   (1.067)    (0.098)
Taxes                        0.729     0.809     -0.080
                            (0.643)   (0.514)    (0.050)
Sales Taxes                  0.427     0.386      0.041
                            (0.357)   (0.266)    (0.027)
Property Taxes               0.217     0.327     -0.110 (***)
                            (0.236)   (0.339)    (0.026)
Charges and Misc. Revenue    0.558     0.491      0.067
                            (0.594)   (0.332)    (0.040)
Financial Transactions       0.085     0.074      0.011
                            (0.336)   (0.136)    (0.021)
State Intergov. Revenue      0.174     0.207     -0.033
                            (0.235)   (0.300)    (0.024)
Total Expenditures           1.947     2.035     -0.088
                            (1.446)   (1.321)    (0.120)
Police Spending              0.214     0.240     -0.026 (**)
                            (0.100)   (0.088)    (0.008)
Capital Outlays              0.413     0.420     -0.007
                            (0.394)   (0.329)    (0.031)
Financial Administration     0.039     0.041     -0.002
                            (0.038)   (0.036)    (0.003)
Cash Securities              2.440     2.672     -0.232
                            (7.841)   (4.037)    (0.519)
Total Debt Outstanding       2.620     2.682     -0.062
                            (7.686)   (4.035)    (0.512)
Within-City Std. Dev.        0.059     0.043      0.016 (**)
Sales Tax                   (0.064)   (0.042)    (0.005)
Within-City Std. Dev.        0.209     0.215     -0.006
Own-Source Revenue          (0.184)   (0.177)    (0.017)
Within-City Std. Dev.        0.247     0.269     -0.022
Total Expenditures          (0.281)   (0.267)    (0.026)
Observations               228       310        538

NOTE: This table reports summary statistics for variables used in this
paper. It reports differences in means for vairables used on a per
capita basis between municipalities with a bancrupt chain and those
without a bankrupt chain for 2005 through 2007. For per-capita
analysis, we remove nine cities with populations below 1,000. This is
not important in other analyses since we account for city fixed
effects, but for per-capita analysis, these outliers drastically change
the means. mean coefficients; standard deviations in parentheses.

Table 3: Effect of Bankruptcy on Sales Tax and Gross Receipts Revenue

Panel A                  Sales Tax and Gross Receipt Revenue
                    (1)              (2)              (3)

Bankrupt Dummy        -0.1575 (***)                     -0.0981 (***)
                      (0.0383)                          (0.0291)
Bankrupt Count                         -0.0307 (***)
                                       (0.0088)
State-Year FEs      NO               NO               YES
Adjusted [R.sup.2]     0.947            0.947            0.969
Observations        4346             4346             4346
Panel B                          Own-Source Revenue
                      (1)            (2)                (3)
Bankrupt Dummy        -0.0499 (***)                     -0.0405 (**)
                      (0.0142)                          (0.0174)
Bankrupt Count                         -0.0186 (***)
                                       (0.0040)
State-Year FEs      NO               NO               YES
Adjusted [R.sup.2]     0.988            0.988            0.989
Observations        4350             4350             4350

Panel A             Sales Tax and Gross Receipt Revenue
                    (4)

Bankrupt Dummy
Bankrupt Count        -0.0167 (**)
                      (0.0054)
State-Year FEs      YES
Adjusted [R.sup.2]     0.969
Observations        4346
Panel B             Own-Source Revenue
                    (4)
Bankrupt Dummy
Bankrupt Count        -0.0180 (***)
                      (0.0054)
State-Year FEs      YES
Adjusted [R.sup.2]     0.989
Observations        4350

Note: This table reports estimates of regressions of the following form
for the pooled bankruptcy sample:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

where [Revenue.sub.it] is sales tax and gross receipt revenue in Panel
A and own-source revenue in Panel B in municipality i, in year t;
[BankruptDummy.sub.i] equals 1 when municipality i contains one or more
of the treatment chains of type c. [Post.sub.t] is an indicator for
whether or not the year is after 2008, the bankruptcy year. Operating
Dummy * Post controls for whether or not municipality i contains any
operating stores after the bankruptcy year. [[delta].sub.i] represents
municipality fixed effects and [[gamma].sub.t] represents year fixed
effects. Columns 2 and 4 of this table replace [BankruptDummy.sub.i]
with [BankruptCount.sub.i], which is the number of bankrupt big-box
stores in municipality i. State-by-year fixed effects are included in
columns 3 and 4. Standard errors clustered at the county level are in
parentheses. Municipalities are excluded if they have over 50 big-box
stores, have a 500% change between their maximum total revenue (or
sales tax or total expenditure) and their minimum total revenue (or
sales tax or total expenditure), or if they are in the data for less
than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table 4: Effect of Bankruptcy on Sales Tax and Own-Source Revenue Over
Time

                              (1)                     (2)
                              Sales Tax and Gross
                              Receipts

One Year After Bankruptcy        -0.1469 (***)           -0.0771 (***)
                                 (0.0379)                (0.0244)
Two Years After Bankruptcy       -0.1501 (***)           -0.1029 (***)
                                 (0.0402)                (0.0339)
Three Years After Bankruptcy     -0.1663 (***)           -0.1086 (***)
                                 (0.0392)                (0.0302)
Four Years After Bankruptcy      -0.1651 (***)           -0.1029 (***)
                                 (0.0405)                (0.0343)
State-Year FEs                NO                      YES
[R.sup.2]                         0.947                   0.969
Observations                  4,346                   4,346

                              (3)               (4)
                              Own-Source Revenue

One Year After Bankruptcy        -0.0803 (***)     -0.0929 (***)
                                 (0.0227)          (0.0232)
Two Years After Bankruptcy       -0.0311 (**)      -0.0301 (*)
                                 (0.0150)          (0.0179)
Three Years After Bankruptcy     -0.0290 (*)       -0.0133
                                 (0.0166)          (0.0203)
Four Years After Bankruptcy      -0.0581 (***)     -0.0278
                                 (0.0183)          (0.0232)
State-Year FEs                NO                YES
[R.sup.2]                         0.988             0.989
Observations                  4,350             4,350

Note: This table reports estimates of regressions of the following form
using the pooled bankruptcy sample:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptDummy *
YearDummy).sup.c.sub.it] + [[theta](OperatingDummy *
YearDummy).sup.c.sub.it] + + [[delta].sub.i] + [[gamma].sub.t]
+[[epsilon].sub.it]

where [Revenue.sub.it] is sales tax and gross receipt revenue for
columns 1 and 2 and own-source revenue in columns 3 and 4 in
municipality i, in year t; [BankruptDummy.sup.c.sub.i] equals 1 when
municipality i contains one or more of the treatment chains of type c,
where the store type is electronics, department store, or both;
[YearDummy.sub.t] represents dummy variables for each of the four years
after the bankruptcy OperatingDummy * YearDummy controls for whether or
not municipality i contains any operating stores in category c during
the corresponding year. [[delta].sub.i] represents municipality fixed
effects and [[gamma].sub.t] represents year fixed effects.
State-by-year fixed effects are included in columns 2 and 4. Standard
errors clustered at the county level are in parentheses. Municipalities
are excluded if they have over 50 big-box stores, have a 500% change
between their maximum total revenue (or sales tax or total expenditure)
and their minimum total revenue (or sales tax or total expenditure), or
if they are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table 5: Effect of Bankruptcy on Spending and the Revenue Recovery
Thereafter

                    Total            Police           Capital
                    Expenditure      Protection       Outlays

Panel A
Bankrupt              -0.0336 (**)     -0.0266 (*)      -0.0274
Dummy                 (0.0164)         (0.0148)         (0.0586)
Adjusted [R.sup.2]     0.986            0.984            0.828
Panel B
Bankrupt              -0.0145 (***)    -0.0076 (***)    -0.0297 (**)
Count                 (0.0040)         (0.0029)         (0.0134)
Adjusted [R.sup.2]     0.986            0.984            0.828
Observations        4350             4348             4329
                    (1)              (2)              (3)
                    Property         Charges &        Fin.
                    Tax              Misc.            Transactions
                                     Revenue
Panel C
Bankrupt               0.0289           0.0061           0.1346 (*)
Dummy                 (0.0417)         (0.0256)         (0.0716)
Adjusted [R.sup.2]     0.965            0.974            0.857
Panel D
Bankrupt              -0.0038           0.0025           0.0196
Count                 (0.0089)         (0.0057)         (0.0183)
Adjusted [R.sup.2]     0.965            0.974            0.857
Observations        4208             4350             4348

                    Financial Ad-    Total Debt    Cash
                    ministration     Outstanding   Securities

Panel A
Bankrupt              -0.1016 (*)      -0.0715       -0.0717 (*)
Dummy                 (0.0532)         (0.0455)      (0.0389)
Adjusted [R.sup.2]     0.827            0.961         0.955
Panel B
Bankrupt              -0.0324 (**)     -0.0125       -0.0208 (**)
Count                 (0.0144)         (0.0094)      (0.0087)
Adjusted [R.sup.2]     0.827            0.960         0.955
Observations        4205             4340          4348
                    (4)              (5)           (6)
                    Property Tax     Property Tax  Property
                    & Fin.           & Charges     Tax &
                    Transactions                   Charges &
                                                   Fin. Transactions

Panel C
Bankrupt               0.0931 (***)     0.0296        0.0319
Dummy                 (0.0324)         (0.0198)      (0.0207)
Adjusted [R.sup.2]     0.966            0.985         0.983
Panel D
Bankrupt               0.0046           0.0020        0.0021
Count                 (0.0083)         (0.0045)      (0.0047)
Adjusted [R.sup.2]     0.965            0.984         0.983
Observations        4349             4350          4350

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([Expenditure.sup.h.sub.it]) = [alpha] + [[beta](BankruptDummy *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] +
[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [Expenditure.sup.h.sub.it] is the amount of local government
expenditures in category h. There are eight categories, which are total
expenditures, police protection, capital outlays, financial
administration, total debt outstanding, and cash securities in
municipality i, in year t. [BankruptDummy.sub.i] equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not the year is after 2008,
the bankruptcy year. Operating Dummy * [Post.sub.it] controls for
whether or not municipality i contains any operating stores after the
bankruptcy year. [[delta].sub.i] represents municipality fixed effects
and [[gamma].sub.t] represents year fixed effects. Panels B and D of
this table replace [BankruptDummy.sub.i] with [BankruptCount.sub.i],
which is the number of bankrupt big-box store in municipality i. Panel
C and Panel D replace [Expenditure.sup.h.sub.it] with
[Revenue.sup.h.sub.it], which is the amount of local government revenue
in category h. There are six categories, which are property taxes;
charges and miscellaneous revenue; financial transactions; property tax
and financial transactions; property tax and charges; and property tax
and charges and financial transactions in municipality i, in year t.
Panel A and Panel C include [BankruptDummy.sub.i], but in Panels B and
D, [BankruptDummy.sub.i] is replaced with [BankruptCount.sub.i].
Standard errors clustered at the county level are in parentheses. This
table uses the pooled sample of municipalities. Municipalities are
excluded if they have over 50 big-box stores, have a 500% change
between their maximum total revenue (or sales tax or total expenditure)
and their minimum total revenue (or sales tax or total expenditure), or
if they are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table 6: Post-Bankruptcy Comparison on Revenue Sources and Total
Expenditures between Home Rule and Non-Home Rule Cities

Panel A                   (1)              (2)
                          Sales Tax &      Own-Source
                          Gross            Rev.
                          Receipts

Bankrupt Dummy              -0.1325          -0.0721 (***)
                            (0.0907)         (0.0265)
Home Rule x Bankrupt        -0.0146           0.0526 (*)
Dummy                       (0.0970)         (0.0309)
Combined Effect:            -0.1471 (***)    -0.0196
Bankrupt+(HR x Bankrupt)    (0.0404)         (0.0185)
Adjusted [R.sup.2]           0.946            0.987
Observations              4337             4341
Panel B                   (1)              (2)
                          Sales Tax &      Own-Source
                          Gross            Rev.
                          Receipts
Bankrupt Count              -0.0668 (**)     -0.0369 (***)
                            (0.0276)         (0.0074)
Home Rule x Bankrupt         0.0324           0.0221 (**)
Count                       (0.0275)         (0.0087)
Combined Effect:            -0.0344 (***)    -0.0149 (**)
Bankrupt+(HR x Bankrupt)    (0.0116)         (0.0063)
Adjusted [R.sup.2]           0.946            0.987
Observations              4337             4341

Panel A                   (3)              (4)
                          Property         Total
                          Taxes,           Expenditures
                          Charges &
                          Fees, & Misc.
                          Rev.

Bankrupt Dummy              -0.0108          -0.0495 (*)
                            (0.0369)         (0.0266)
Home Rule x Bankrupt         0.0736 (*)       0.0393
Dummy                       (0.0428)         (0.0339)
Combined Effect:             0.0628 (***)    -0.0102
Bankrupt+(HR x Bankrupt)    (0.0238)         (0.0217)
Adjusted [R.sup.2]           0.984            0.986
Observations              4341             4341
Panel B                   (3)              (4)
                          Property         Total
                          Taxes,           Expenditures
                          Charges &
                          Fees, & Misc.
                          Rev.
Bankrupt Count              -0.0068          -0.0274 (***)
                            (0.0160)         (0.0093)
Home Rule x Bankrupt         0.0156           0.0130
Count                       (0.0150)         (0.0101)
Combined Effect:             0.0088          -0.0145 (**)
Bankrupt+(HR x Bankrupt)    (0.0059)         (0.0072)
Adjusted [R.sup.2]           0.984            0.986
Observations              4341             4341

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([Outcome.sub.it]) = [alpha]+ [[??](Bankrupt Dummy * Post *
HomeRule).sub.it] + [[beta](BankruptDummy * Post).sub.it] +
[[rho](HomeRule * Post).sub.it] + [[theta](OperatingDummy *
Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [Outcome.sub.it] is the sales tax and gross receipts revenue,
own-source revenue, property taxes, charges and fees, and miscellenious
revenue, and total expenditures in municipality i, in year t in columns
1, 2, 3, and 4 respectively. [BankruptDummy.sub.i] equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not year t is after 2008
(the bankruptcy year). [HomeRule.sub.i] is a dummy variable equal to 1
if the municipality has home rule status according to the ACIR 1993
home rule measure. OperatingDummy * Post controls for any operating
stores after the bankruptcy year. [[delta].sub.i] represents
municipality fixed effects and [[gamma].sub.t] represents time fixed
effects. The "Combined Effect" row shows the sum of the coefficient on
BankruptDummy * Post and the coefficient on the interaction term
HomeRule * BankruptDummy * Post. This gives us the total effect of the
bankruptcy on home rule municipalities. Panel B of this table replaces
[BankruptDummy.sub.i] with [BankruptCount.sub.i], which is the total
number of bankrupt stores in municipality i. Standard errors clustered
at the county level are in parentheses. Municipalities are excluded if
they have over 50 big-box stores, have a 500% change between their
maximum total revenue (or sales tax or total expenditure) and their
minimum total revenue (or sales tax or total expenditure), or if they
are in the data for less than 5 years.

(*) p < 0.1, (**) p< 0.05, (***) p < 0.01

Table 7: Home Rule Regression Discontinuity Analysis

                         (1)                       (2)

Panel A                  First Stage, Municipality has Home Rule
Population[greater than               0.591 (***)    0.592 (***)
or equal to]25,000                   (0.116)        (0.106)
Observations                        148            183
Panel B                  Largest Percent Fall in Revenue from 2010-2015
Municipality has                     -0.100 (**)    -0.081 (*)
Home Rule                            (0.048)        (0.048)
Observations                        148            183
Panel C                  Percent Fall in Revenue Greater Than 10%
Municipality has                     -0.293 (**)    -0.228 (*)
Home Rule                            (0.136)        (0.131)
Observations                        148            183
Panel D                  Percent Fall in Revenue Greater Than 30%
Municipality has                     -0.198 (**)    -0.182 (**)
Home Rule                            (0.092)        (0.090)
Observations                        148            183
Panel E                  Municipality has Extremely Strong Bond Rating
Municipality has                      0.328                     0.330
Home Rule                            (0.253)                   (0.231)
Observations                        283                       331
Clusters                            135                       162
Pop. Bandwidth           [+ or -]12,500            [+ or -]15,000

                         (3)

Panel A                  First Stage, Municipality has Home Rule
Population[greater than               0.578 (***)
or equal to]25,000                   (0.096)
Observations                        259
Panel B                  Largest Percent Fall in Revenue from 2010-2015
Municipality has                     -0.081 (*)
Home Rule                            (0.047)
Observations                        257
Panel C                  Percent Fall in Revenue Greater Than 10%
Municipality has                     -0.197
Home Rule                            (0.129)
Observations                        258
Panel D                  Percent Fall in Revenue Greater Than 30%
Municipality has                     -0.185 (**)
Home Rule                            (0.086)
Observations                        258
Panel E                  Municipality has Extremely Strong Bond Rating
Municipality has                      0.353 (*)
Home Rule                            (0.210)
Observations                        434
Clusters                            218
Pop. Bandwidth           [+ or -]18,000

                         (4)

Panel A                  First Stage, Municipality has Home Rule
Population[greater than               0.594 (***)
or equal to]25,000                   (0.090)
Observations                        314
Panel B                  Largest Percent Fall in Revenue from 2010-2015
Municipality has                     -0.085 (*)
Home Rule                            (0.044)
Observations                        312
Panel C                  Percent Fall in Revenue Greater Than 10%
Municipality has                     -0.177
Home Rule                            (0.124)
Observations                        313
Panel D                  Percent Fall in Revenue Greater Than 30%
Municipality has                     -0.192 (**)
Home Rule                            (0.081)
Observations                        313
Panel E                  Municipality has Extremely Strong Bond Rating
Municipality has                      0.346 (*)
Home Rule                            (0.193)
Observations                        498
Clusters                            257
Pop. Bandwidth           [+ or -]20,000

Note: Panel A of this table reports estimates of first-stage
regressions of the following form:

[HomeRule.sub.i] = [alpha] + [[beta](Above25000).sub.i] +
[theta][(Population).sub.i] + [[rho](Above25000 * Population).sub.i] +
[[epsilon].sub.i]

where [HomeRule.sub.i] is a dummy variable equal to 1 if municipality i
ever had home rule status between 2010 and 2015. [Above25000.sub.i] is
a dummy variable that equals to 1 if municipality's population exceeded
25,000 and equals to 0 otherwise. Population is the maximum number of
population municipality i had sometime between 1994 and 2009.
[Above25000.sub.*] Population is an interaction variable between
Above25000 and Population. This regression establishes a link between
the home rule population threshold in Illinois and a city's actual home
rule status. Panels B, C, D, and E show the results of regressions with
fuzzy regression discontinuity design, using instrumented HomeRule
variable to estimate various public-finance-related variables. Panel B
reports estimates of regressions of the following form:

[RevFall.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] + [lambda][(HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall.sub.i] is the largest annual percentage fall in revenue
from 2010 to 2015 in municipality i. Panel C reports estimates of
regressions of the following form:

[RevFall10.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] + [[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall10.sub.i] is a dummy variable that equals to1 if
municipality i has a fall in revenue larger than 10% at any point from
2010-2015. Panel D reports estimates of regressions of the following
form:

[RevFall30.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] +[[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall30.sub.i] is a dummy variable that equals to 1 if
municipality i has a fall in revenue larger than 30% at any point from
2010-2015. Panel E reports estimates of regressions of the following
form:

[StrongBond.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] + [[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [StrongBond.sub.i] is a dummy variable that equals to 1 if
municipality i has extremely strong bond rating in IL data (from 1994
to 1996) or in scraped data (2015).

Regressions from all five panels are run with four different population
bandwidths. Column 1 includes cities with populations between 12,500
and 37,500; column 2 includes cities with populations between 10,000
and 40,000; column 3 includes cities with populations between 7,000 and
43,000; and column 4 includes cities with populations between 5,000 and
45,000. Standard errors clustered at the municipality level are in
parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.1: Main Results are Robust to Sample Restriction

                    (1)                (2)
                    Sales Tax & Gross  Own-Source
                    Receipts Revenue   Revenue

Panel A             Cities in Data Three or More Years
Bankrupt Dummy        -0.1582 (***)      -0.0488 (***)
                      (0.0381)           (0.0142)
Observations        4363               4367
Adjusted [R.sup.2]     0.948              0.988
Panel B             Cities in Data Four or More Years
Bankrupt Dummy        -0.1582 (***)      -0.0488 (***)
                      (0.0381)           (0.0142)
Observations        4363               4367
Adjusted [R.sup.2]     0.948              0.988
Panel C             Cities in Data Six or More Years
Bankrupt Dummy        -0.1599 (***)      -0.0528 (***)
                      (0.0408)           (0.0150)
Observations        3928               3929
Adjusted [R.sup.2]     0.944              0.987
Panel D             Cities in Data Seven or More Years
Bankrupt Dummy        -0.1598 (***)      -0.0527 (***)
                      (0.0422)           (0.0155)
Observations        3712               3713
Adjusted [R.sup.2]     0.941              0.986

                    (3)
                    Total Expenditures

Panel A             Cities in Data Three or More Years
Bankrupt Dummy        -0.0330 (**)
                      (0.0164)
Observations        4367
Adjusted [R.sup.2]     0.986
Panel B             Cities in Data Four or More Years
Bankrupt Dummy        -0.0330 (**)
                      (0.0164)
Observations        4367
Adjusted [R.sup.2]     0.986
Panel C             Cities in Data Six or More Years
Bankrupt Dummy        -0.0349 (**)
                      (0.0172)
Observations        3929
Adjusted [R.sup.2]     0.986
Panel D             Cities in Data Seven or More Years
Bankrupt Dummy        -0.0317 (*)
                      (0.0176)
Observations        3713
Adjusted [R.sup.2]     0.985

Note: Typically, we allow municipalities in our sample if they are in
the data for 5 or more years. In this table, we change that sample
restriction as a robustness check. In all panels, we run the same
regressions listed in Tables 2 and 3. In Panel A, we allow
municipalities in our sample if they are in the data for 3 or more
years. In Panel B, we allow municipalities in our sample if they are in
the data for 4 or more years. In Panel C, we allow municipalities in
our sample if they are in the data for 6 or more years. In Panel D, we
allow municipalities in our sample if they are in the data for 7 or
more years. This sample restriction does not affect the results.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.2: Summary Statistics by Home Rule Status

                        Krane et al. (2002)
                        State-Level Measure
                        Not Home Rule  Home Rule

Population               109,659          121,944
                        (191,603)        (203,985)
Number of Big-Box              2.759            3.541
                              (2.689)          (3.461)
Total Revenue            335,228          282,942
                        (928,556)        (691,208)
Own-Source Revenue       274,666          236,797
                        (667,519)        (561,002)
Taxes                    109,427          101,092
                        (313,337)        (208,953)
Total Expenditures       327,327          290,631
                        (885,429)        (695,875)
Total Debt Outstanding   337,014          483,514
                        (896,226)      (1,450,430)
Observations                 783             3558

                        ICMA (1974)
                        Municipality-Level Measure
                        Not Home Rule    Home Rule

Population                 125,610          142,936
                          (248,466)        (196,893)
Number of Big-Box                3.054            4.438
                                (2.126)          (4.558)
Total Revenue              301,118          317,511
                          (754,940)        (563,074)
Own-Source Revenue         250,754          275,638
                          (633,135)        (469,592)
Taxes                      109,049          102,044
                          (224,677)        (148,985)
Total Expenditures         310,513          325,852
                          (793,521)        (577,036)
Total Debt Outstanding     464,548          540,872
                        (1,722,903)      (1,284,031)
Observations                  1326             1413

This table shows summary statistics for the municipalities in our
sample based on whether or not they are defined as "home rule"
municipalities. In the first two columns, the "home rule" breakdown is
based on a state-level measure from Krane et al. (2002). In the last
two columns, the "home rule" breakdown is based on a municipality-level
measure from the 1974 International City/County Management Association
(ICMA) survey. Standard deviations are in parentheses. The statistics
provided are mean and standard deviation of municipality's population,
the number of big-box stores (the number of Best Buy, Circuit City,
CompUSA, Mervyns, Kohls, and JC Penney stores), municipality's total
revenue, own-source revenue, total taxes, total spending, and total
debt outstanding.

Table A.3: Summary Statistics by Municipality Population Category

                                   Municipality Population Category
                                    1,000-5,000        5,000-45,000

Panel A                            Municipality Finance Statistics
Per Capita Revenue                  1,723.705          2,199.596
                                   (2,484.743)        (2,085.808)
Per Capita Own-Source Revenue       1,345.909          1,781.923
                                   (2,287.647)        (1,994.934)
Per Capita Local Sales Tax              5.008             42.259
                                      (52.630)          (120.489)
Per Capita Property Tax               219.788            301.350
                                     (585.384)          (246.540)
Per Capita Charges and Fees           503.115            574.656
                                     (779.569)          (927.799)
Per Capita Financial Transactions      11.003             55.182
                                      (41.129)           (70.010)
Home Rule Status                        0.122              0.409
                                       (0.328)            (0.492)
Annual Percent Change in Per            0.032              0.032
Capita Revenue                         (0.246)            (0.133)
Observations                         1797               1503
Unit of Observation                Municipality-Year  Municipality-Year
Panel B                            Revenue Stability Statistics
Rev. Fall                               0.199              0.098
                                       (0.263)            (0.143)
Per Capita Total Revenue in 2010    1,572.405          1,985.039
                                   (2,437.439)        (1,538.591)
Home Rule Status                        0.127              0.415
                                       (0.334)            (0.493)
Rev. Fall > 10 %                        0.558              0.302
                                       (0.497)            (0.460)
Rev. Fall > 30 %                        0.193              0.071
                                       (0.395)            (0.257)
Observations                          362                311
Unit of Observation                Municipality       Municipality
Panel C                            Bond Rating Statistics
Extremely Strong                        0.062              0.168
                                       (0.242)            (0.374)
Very Strong                             0.048              0.164
                                       (0.215)            (0.371)
Strong                                  0.166              0.352
                                       (0.373)            (0.478)
Adequate or Less                        0.097              0.192
                                       (0.296)            (0.394)
Missing Bond Rating                     0.697              0.212
                                       (0.461)            (0.409)
Home Rule Status                        0.207              0.496
                                       (0.406)            (0.500)
Observations                          145                500
Unit of Observation                Municipality-Year  Municipality-Year

                                   Municipality Population Category
                                   45,000-200,000

Panel A
Per Capita Revenue                  2,760.967
                                   (1,123.868)
Per Capita Own-Source Revenue       2,276.569
                                   (1,078.800)
Per Capita Local Sales Tax            112.063
                                      (78.475)
Per Capita Property Tax               410.851
                                     (156.184)
Per Capita Charges and Fees           637.184
                                     (494.625)
Per Capita Financial Transactions     106.967
                                      (76.563)
Home Rule Status                        0.973
                                       (0.164)
Annual Percent Change in Per            0.030
Capita Revenue                         (0.121)
Observations                          146
Unit of Observation                Municipality-Year
Panel B                            Revenue Stability Statistics
Rev. Fall                              0.089
                                      (0.117)
Per Capita Total Revenue in 2010   2,489.151
                                    (982.864)
Home Rule Status                       0.970
                                      (0.174)
Rev. Fall > 10 %                       0.273
                                     i(0.452)
Rev. Fall > 30 %                       0.061
                                      (0.242)
Observations                          33
Unit of Observation                Municipality
Panel C                            Bond Rating Statistics
Extremely Strong                       0.113
                                      (0.318)
Very Strong                            0.425
                                      (0.497)
Strong                                 0.412
                                      (0.495)
Adequate or Less                       0.075
                                      (0.265)
Missing Bond Rating                    0.000
                                      (0.000)
Home Rule Status                       0.950
                                      (0.219)
Observations                          80
Unit of Observation                Municipality-Year

This table reports summary statistics for our sample of municipalities
by population category. Panel A shows summary statistics variables
relating to municipal taxes and revenue at the municipality-year
observation level. Panel B shows summary statistics for measures of
revenue stability at the municipality level. Finally, Panel C Shows
summary statistics for variables relating to municipal bond ratings at
the municipality-year level. It is important to note that observations
with missing values for bond ratings were coded as 0 for Extremel
yStrong. The probability of a missing value does not change around the
cutoffs. For all panels, mean coefficients are presented, with standard
deviations in parentheses.

(*) p < 0.05, (**) p < 0.01, (***) p < 0.001

Table A.4: Main Results are Robust to Covariates and Inclusion of
County Finances

                    (1)                (2)              (3)
                    Sales Tax & Gross  Own-Source       Total
                    Receipts Revenue   Revenue          Expenditures

Panel A             Omitting Controls for Operating Stores
Bankrupt Dummy        -0.1467 (***)      -0.0447 (***)    -0.0336 (**)
                      (0.0386)           (0.0145)         (0.0164)
Observations        4346               4350             4350
Adjusted [R.sup.2]     0.947              0.988            0.986
Panel B             Controls for State-Level Finances and Unemployment
Bankrupt Dummy        -0.1269 (***)      -0.0480 (***)    -0.0302 (*)
                      (0.0405)           (0.0158)         (0.0166)
Observations        4346               4350             4350
Adjusted [R.sup.2]     0.947              0.988            0.986
Panel C             Controls for County-Level Finances
Bankrupt Dummy        -0.0933 (***)      -0.0692 (***)    -0.0394 (**)
                      (0.0290)           (0.0180)         (0.0181)
Observations        3175               3174             3175
Adjusted [R.sup.2]     0.975              0.986            0.988

Note: In Panel A, we run the same regressions listed in Tables 2 and 3
but without the OperatingDummy*Post term. This regression takes the
following form:

ln([Outcome.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

Column 1 shows this result for sales tax and gross receipt revenue,
column 2 shows this for own-source revenue, and column 3 shows this for
total expenditures. The inclusion or exclusion of this covariate does
not affect the results. BankruptDummyi takes a value of 1 if a
municipality has a bankrupt chain and OperatingDummyi if there is any
operating chain in that municipality. These are both interacted with
[Post.sub.t] which equals 1 after 2008. In Panel B, we run the same
regressions listed in Tables 2 and 3 but with state-level finance and
unemployment rate controls added (these vary at the state-year level).
This regression takes the following form:

ln([Outcome.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[xi].sub.it] +
[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

The specific controls state-level controls we include in
[[zeta].sub.it] are: total revenue, total taxes, total expenditures,
and total debt outstanding (from Census of Local Government Finance)
and annual unemployment rate (from BLS Local Area Unemployment
Statistics and Current Population Survey). Column 1 shows this result
for sales tax and gross receipt revenue, column 2 shows this for
own-source revenue, and column 3 shows this for total expenditures. The
inclusion or exclusion of this covariate does not affect the results.
In Panel C, we run the same regressions listed in Panel B but wothout
the state control vector and on data that is aggregated to the
county-level instead of the municipality-level. Now, the finance
measures (sales tax, own-source revenue, and total expenditures)
include finances at township, municipality, and county level all
aggregated to county as the unit of observation. This does not include
school district finances or other special district finances since
reporting does not appear to be as consistent from year to year. Column
1 shows this result for sales tax and gross receipt revenue, column 2
shows this for own-source revenue, and column 3 shows this for total
expenditures. Including county-level finances does not affect the
results.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.5: Relationship between Big-Box Bankruptcy and
Municipality-Level Population Estimates

                                     Population
                    (1)              (2)              (3)

Bankrupt Dummy        -0.0076                           -0.0059
                      (0.0055)                          (0.0066)
Bankrupt Count                         -0.0020
                                       (0.0014)
Constant              11.1121 (***)    11.1121 (***)    11.1114 (***)
                      (0.0048)         (0.0049)         (0.0037)
State-Year FEs      NO               NO               YES
Observations        4263             4263             4263
Adjusted [R.sup.2]     0.999            0.999            0.999

                    Population
                    (4)

Bankrupt Dummy
Bankrupt Count        -0.0037 (*)
                      (0.0021)
Constant              11.1113 (***)
                      (0.0037)
State-Year FEs      YES
Observations        4263
Adjusted [R.sup.2]     0.999

Note: This table reports estimates of regressions of the following
form:

ln([Population.sub.it]) = [alpha] + [[beta](BankruptDummy *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] +
[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [Population.sub.it] is the U.S. Census Bureau population estimate
in municipality i, in year t. [BankruptDummy.sub.i] equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not the year is after 2008,
the bankruptcy year. OperatingDummy * [Post.sub.it] controls for
whether or not municipality i contains any operating stores after the
bankruptcy year. [[delta].sub.i] represents municipality fixed effects
and [[gamma].sub.t] represents year fixed effects. Columns 2 and 4 of
this table replace [BankruptDummy.sub.i] with [BankruptCount.sub.i],
which is the number of bankrupt big-box store in municipality i.
Columns 1 and 2 show estimates without state-by-year fixed effects and
columns 3 and 4 include state-by-year fixed effects. Standard errors
clustered at the county level are in parentheses. This table uses U.S.
Census Bureau estimates for municipality population. The U.S. Census
Bureau estimates county population in each year by using administrative
records on county level births, deaths, and migration. This
county-level estimate is then applied to municipalities based on
existing housing unit counts at the sub-county level. For the following
analysis, we remove cities that are extreme outliers in terms of
min-to-max population change from 2004-2012. Additionally,
municipalities are excluded if they have over 50 big-box stores, have a
500% change between their maximum total revenue (or sales tax or total
expenditure) and their minimum total revenue (or sales tax or total
expenditure), or if they are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.6: Effect of Big-Box Bankruptcy on Sales Tax and Gross
Receipts, Including Population Controls

                         Sales Tax and Gross Receipts
                    (1)              (2)              (3)

Bankrupt Dummy        -0.1518 (***)                     -0.1033 (***)
                      (0.0388)                          (0.0306)
Bankrupt Count                         -0.0378 (***)
                                       (0.0087)
Constant               9.6386 (***)     9.5284 (***)     9.7346 (***)
                      (0.1234)         (0.1331)         (0.1001)
State-Year FEs      NO               NO               YES
Observations        4259             4259             4259
Adjusted [R.sup.2]     0.946            0.945            0.969

                    Sales Tax and Gross Receipts
                    (4)

Bankrupt Dummy
Bankrupt Count        -0.0195 (***)
                      (0.0062)
Constant               9.6953 (***)
                      (0.1141)
State-Year FEs      YES
Observations        4259
Adjusted [R.sup.2]     0.969

Note: This table reports estimates of regressions of the following
form:

ln([SalesTax.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] +[pi]([Population.sub.it]) +
[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [SalesTax.sub.it] is the sales tax and gross receipt revenue in
municipality i, in year t. [BankruptDummy.sub.i] equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not the year is after 2008,
the bankruptcy year. OperatingDummy * [Post.sub.it] controls for
whether or not municipality i contains any operating stores after the
bankruptcy year. [Population.sub.it] is the U.S. Census Bureau
population estimate for municipality i in year t. [[delta].sub.i]
represents municipality fixed effects and [[gamma].sub.t] represents
year fixed effects. Columns 2 and 4 of this table replace
[BankruptDummy.sub.i] with [BankruptCount.sub.i], which is the number
of bankrupt big-box store in municipality i. Columns 1 and 2 show
estimates without state-by-year fixed effects and columns 3 and 4
include state-by-year fixed effects. Standard errors clustered at the
county level are in parentheses. This table uses U.S. Census Bureau
estimates for municipality population. The U.S. Census Bureau estimates
county population in each year by using administrative records on
county level births, deaths, and migration. This county-level estimate
is then applied to municipalities based on existing housing unit counts
at the sub-county level. For the following analysis, we remove cities
that are extreme outliers in terms of min-to-max population change from
2004-2012. Additionally, municipalities are excluded if they have over
50 big-box stores, have a 500% change between their maximum total
revenue (or sales tax or total expenditure) and their minimum total
revenue (or sales tax or total expenditure), or if they are in the data
for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.7: Effect of Big-Box Bankruptcy on Own-Source Revenue,
Including Population Controls

                             Own-Source Revenue
                    (1)              (2)              (3)

Bankrupt Dummy        -0.0467 (***)                     -0.0349 (*)
                      (0.0149)                          (0.0179)
Bankrupt Count                         -0.0223 (***)
                                       (0.0053)
Constant              11.3306 (***)    11.2319 (***)    11.4026 (***)
                      (0.0731)         (0.0820)         (0.0727)
State-Year FEs      NO               NO               YES
Observations        4263             4263             4263
Adjusted [R.sup.2]     0.987            0.988            0.989

                    Own-Source Revenue
                    (4)

Bankrupt Dummy
Bankrupt Count        -0.0199 (***)
                      (0.0069)
Constant              11.3097 (***)
                      (0.0864)
State-Year FEs      YES
Observations        4263
Adjusted [R.sup.2]     0.989

Note: This table reports estimates of regressions of the following
form:

ln([OwnSourceRev.sub.it]) = [alpha] + [[beta](BankruptDummy *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] +
[pi]([Population.sub.it]) + [[delta].sub.i] + [[gamma].sub.t] +
[[epsilon].sub.it]

where [OwnSource.sub.it] is the own-source revenue in municipality i,
in year t. [BankruptDummy.sub.i] equals 1 when municipality i contains
one or more of the treatment chains. [Post.sub.t] is an indicator for
whether or not the year is after 2008, the bankruptcy year.
OperatingDummy * [Post.sub.it] controls for whether or not municipality
i contains any operating stores after the bankruptcy year.
[Population.sub.it] is the U.S. Census Bureau population estimate for
municipality i in year t. [[delta].sub.i] represents municipality fixed
effects and [[gamma].sub.t] represents year fixed effects. Columns 2
and 4 of this table replace [BankruptDummy.sub.i] with
[BankruptCount.sub.i], which is the number of bankrupt big-box store in
municipality i. Columns 1 and 2 show estimates without state-by-year
fixed effects and columns 3 and 4 include state-by-year fixed effects.
Standard errors clustered at the county level are in parentheses. This
table uses U.S. Census Bureau estimates for municipality population.
The U.S. Census Bureau estimates county population in each year by
using administrative records on county level births, deaths, and
migration. This county-level estimate is then applied to municipalities
based on existing housing unit counts at the sub-county level. For the
following analysis, we remove cities that are extreme outliers in terms
of min-to-max population change from 2004-2012. Additionally,
municipalities are excluded if they have over 50 big-box stores, have a
500% change between their maximum total revenue (or sales tax or total
expenditure) and their minimum total revenue (or sales tax or total
expenditure), or if they are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.8: Effect of Big-Box Bankruptcy on Total Expenditures,
Including Population Controls

                                Total Expenditures
                    (1)              (2)              (3)

Bankrupt Dummy        -0.0349 (**)                      -0.0278
                      (0.0164)                          (0.0172)
Bankrupt Count                         -0.0192 (***)
                                       (0.0039)
Constant              11.4058 (***)    11.3162 (***)    11.4852 (***)
                      (0.0675)         (0.0681)         (0.0734)
State-Year FEs      NO               NO               YES
Observations        4263             4263             4263
Adjusted [R.sup.2]     0.986            0.986            0.987

                    Total Expenditures
                    (4)

Bankrupt Dummy
Bankrupt Count        -0.0177 (***)
                      (0.0044)
Constant              11.3998 (***)
                      (0.0731)
State-Year FEs      YES
Observations        4263
Adjusted [R.sup.2]     0.987

Note: This table reports estimates of regressions of the following
form:

ln([Expenditure.sub.it]) = [alpha] + [[beta](BankruptDummy *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it]
+[pi]([Population.sub.it]) + [[delta].sub.i] + [[gamma].sub.t] +
[[epsilon].sub.it]

where [Expenditure.sub.it] is the total expenditures in municipality i,
in year t. [BankruptDummy.sub.i] equals 1 when municipality i contains
one or more of the treatment chains. [Post.sub.t] is an indicator for
whether or not the year is after 2008, the bankruptcy year. Operating
Dummy * [Post.sub.it] controls for whether or not municipality i
contains any operating stores after the bankruptcy year.
[Population.sub.it] is the U.S. Census Bureau population estimate for
municipality i in year t. [[delta].sub.i] represents municipality fixed
effects and [[gamma].sub.t] represents year fixed effects. Columns 2
and 4 of this table replace [BankruptDummy.sub.i] with
[BankruptCount.sub.i], which is the number of bankrupt big-box store in
municipality i. Columns 1 and 2 show estimates without state-by-year
fixed effects and columns 3 and 4 include state-by-year fixed effects.
Standard errors clustered at the county level are in parentheses. This
table uses U.S. Census Bureau estimates for municipality population.
The U.S. Census Bureau estimates county population in each year by
using administrative records on county level births, deaths, and
migration. This county-level estimate is then applied to municipalities
based on existing housing unit counts at the sub-county level. For the
following analysis, we remove cities that are extreme outliers in terms
of min-to-max population change from 2004-2012. Additionally,
municipalities are excluded if they have over 50 big-box stores, have a
500% change between their maximum total revenue (or sales tax or total
expenditure) and their minimum total revenue (or sales tax or total
expenditure), or if they are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.9: Post-Bankruptcy Comparison on Sales Tax and Gross Receipts
between Home Rule and NonHome Rule Cities

Panel A                   Sales Tax and Gross Receipts
                          (1)              (2)

Bankrupt Dummy              -0.1494 (***)    -0.1325
                            (0.0421)         (0.0907)
ACIR 1993 Home Rule x                        -0.0146
Bankrupt Dummy                               (0.0970)
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:                             -0.1471 (***)
Bankrupt+(HR x Bankrupt)                     (0.0404)
Adjusted [R.sup.2]           0.947            0.946
Observations              4346             4337
Panel B                   Sales Tax and Gross Receipts
                          (1)              (2)
Bankrupt Count              -0.0418 (***)    -0.0668 (**)
                            (0.0118)         (0.0276)
ACIR 1993 Home Rule x                         0.0324
Bankrupt Count                               (0.0275)
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:                             -0.0344 (***)
Bankrupt+(HR x Bankrupt)                     (0.0116)
Adjusted [R.sup.2]           0.947            0.946
Observations              4346             4337

Panel A                   Sales Tax and Gross Receipts
                          (3)              (4)

Bankrupt Dummy              -0.1415          -0.1546
                            (0.1331)         (0.1327)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home       0.0056
Rule 1 x Bankrupt Dummy     (0.1352)
Krane et al. 2002 Home                        0.0211
Rule 2 x Bankrupt Dummy                      (0.1343)
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:            -0.1360 (***)    -0.1335 (***)
Bankrupt+(HR x Bankrupt)    (0.0358)         (0.0355)
Adjusted [R.sup.2]           0.947            0.946
Observations              4337             4337
Panel B                   Sales Tax and Gross Receipts
                          (3)              (4)
Bankrupt Count              -0.0741 (**)     -0.0732 (**)
                            (0.0345)         (0.0343)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home       0.0396
Rule 1 x Bankrupt Count     (0.0345)
Krane et al. 2002 Home                        0.0385
Rule 2 x Bankrupt Count                      (0.0343)
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:            -0.0344 (***)    -0.0347 (***)
Bankrupt+(HR x Bankrupt)    (0.0113)         (0.0113)
Adjusted [R.sup.2]           0.946            0.946
Observations              4337             4337

Panel A                   Sales Tax and Gross Receipts
                          (5)

Bankrupt Dummy              -0.2077 (*)
                            (0.1137)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x        0.0982
Bankrupt Dummy              (0.1257)
Combined Effect:            -0.1095 (**)
Bankrupt+(HR x Bankrupt)    (0.0551)
Adjusted [R.sup.2]           0.928
Observations              2739
Panel B                   Sales Tax and Gross Receipts
                          (5)
Bankrupt Count              -0.0762 (**)
                            (0.0331)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x        0.0549
Bankrupt Count              (0.0353)
Combined Effect:            -0.0213 (*)
Bankrupt+(HR x Bankrupt)    (0.0115)
Adjusted [R.sup.2]           0.927
Observations              2739

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([SalesTax.sub.it]) = [alpha] + [[??] (BankruptDummy * Post *
HomeRule).sup.r.sub.it] + [[beta](BankruptDummy * Post).sub.it]+
[[rho](HomeRule * Post).sup.r.sub.it] + [[theta](OperatingDummy *
Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [SalesTax.sub.it] is the sales tax and gross receipts revenue in
municipality i, in year t. [BankruptDummy.sub.i] equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not year t is after 2008
(the bankruptcy year). [HomeRule.sup.r.sub.i] is a dummy variable equal
to 1 if the municipality has home rule status according to measure r,
where r is one of the four home rule measures discussed in section 3.
OperatingDummy * Post controls for any operating stores after the
bankruptcy year. [[delta].sub.i] represents municipality fixed effects
and [[gamma].sub.t] represents time fixed effects. The "Combined
Effect" row shows the sum of the coefficient on BankruptDummy * Post
and the coefficient on the interaction term HomeRule * BankruptDummy *
Post. This gives us the total effect of the bankruptcy on home rule
municipalities. Panel B of this table replaces [BankruptDummy.sub.i]
with [BankruptCount.sub.i], which is the total number of bankrupt
stores in municipality i. Standard errors clustered at the county level
are in parentheses. Municipalities are excluded if they have over 50
big-box stores, have a 500% change between their maximum total revenue
(or sales tax or total expenditure) and their minimum total revenue (or
sales tax or total expenditure), or if they are in the data for less
than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.10: Post-Bankruptcy Comparison on Own-Source Revenue between
Home Rule and Non-Home Rule Cities

Panel A                      Own-Source Revenue
                             (1)           (2)

Bankrupt Dummy              -0.0368 (**)     -0.0721 (***)
                            (0.0160)         (0.0265)
ACIR 1993 Home Rule x                         0.0526 (*)
Bankrupt Dummy                               (0.0309)
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:                             -0.0196
Bankrupt+(HR x Bankrupt)                     (0.0185)
Adjusted [R.sup.2]           0.988            0.987
Observations              4350             4341
Panel B                         Own-Source Revenue
                            (1)            (2)
Bankrupt Count              -0.0191 (***)    -0.0369 (***)
                            (0.0058)         (0.0074)
ACIR 1993 Home Rule x                         0.0221 (**)
Bankrupt Count                               (0.0087)
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:                             -0.0149 (**)
Bankrupt+(HR x Bankrupt)                     (0.0063)
Adjusted [R.sup.2]           0.988            0.987
Observations              4350             4341

Panel A                       Own-Source Revenue
                               (3)            (4)

Bankrupt Dummy              -0.1149 (***)    -0.1177 (***)
                            (0.0321)         (0.0322)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home       0.1015 (***)
Rule 1 x Bankrupt Dummy     (0.0354)
Krane et al. 2002 Home                        0.1049 (***)
Rule 2 x Bankrupt Dummy                      (0.0351)
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:            -0.0134          -0.0128
Bankrupt+(HR x Bankrupt)    (0.0170)         (0.0169)
Adjusted [R.sup.2]           0.987            0.987
Observations              4341             4341
Panel B                        Own-Source Revenue
                            (3)            (4)
Bankrupt Count             -0.0407 (***)    -0.0401 (***)
                           (0.0081)         (0.0081)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home      0.0257 (***)
Rule 1 x Bankrupt Count    (0.0089)
Krane et al. 2002 Home                       0.0250 (***)
Rule 2 x Bankrupt Count                      (0.0088)
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:           -0.0150 (**)     -0.0152 (**)
Bankrupt+(HR x Bankrupt)   (0.0061)         (0.0061)
Adjusted [R.sup.2]          0.987            0.987
Observations             4341             4341

Panel A                     Own-Source Revenue
                             (5)

Bankrupt Dummy               -0.0561 (**)
                             (0.0231)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x         0.0486
Bankrupt Dummy               (0.0327)
Combined Effect:             -0.0075
Bankrupt+(HR x Bankrupt)     (0.0254)
Adjusted [R.sup.2]            0.987
Observations               2739
Panel B                      Own-Source Revenue
                            (5)
Bankrupt Count               -0.0312 (***)
                             (0.0101)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x         0.0207
Bankrupt Count               (0.0127)
Combined Effect:             -0.0105
Bankrupt+(HR x Bankrupt)     (0.0079)
Adjusted [R.sup.2]            0.987
Observations               2739

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([OwnSourceRev.sub.it]) = [alpha]+ [[??](BankruptDummy * Post *
HomeRule).sup.r.sub.it] + [[beta] (Bankrupt Dummy * Post).sub.it] +
[[rho](HomeRule * Post).sup.r.sub.it] + [[theta](OperatingDummy *
Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [OwnSourceRev.sub.it] is own-source revenue in municipality i, in
year t. [BankruptDummy.sub.i] equals 1 when municipality i contains one
or more of the treatment chains. [Post.sub.t] is an indicator for
whether or not year t is after 2008 (the bankruptcy year).
[HomeRule.sup.r.sub.i] is a dummy variable equal to 1 if the
municipality has home rule status according to measure r, where r is
one of the four home rule measures discussed in section 3.
OperatingDummy * Post controls any operating stores after the
bankruptcy year. [[delta].sub.i] represents municipality fixed effects
and [[gamma].sub.t] represents time fixed effects. The "Combined
Effect" row shows the sum of the coefficient on BankruptDummy * Post
and the coefficient on the interaction term
HomeRule*BankruptDummy*Post. This gives us the total effect of the
bankruptcy on home rule municipalities. Panel B of this table replaces
[BankruptDummy.sub.i] with [BankruptCount.sub.i], which is the total
number of bankrupt stores in municipality i. Standard errors clustered
at the county level are in parentheses. Municipalities are excluded if
they have over 50 big-box stores, have a 500% change between their
maximum total revenue (or sales tax or total expenditure) and their
minimum total revenue (or sales tax or total expenditure), or if they
are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.11: Post-Bankruptcy Comparison on Property Taxes, Charges and
Fees, and Miscellaneous Revenue between Home Rule and Non-Home Rule
Cities

Panel A                       Property Taxes, Charges and Fees, and
                              Misc. Rev.
                              (1)           (2)
Bankrupt Dummy               0.0384 (*)    -0.0108
                            (0.0205)       (0.0369)
ACIR 1993 Home Rule x                       0.0736 (*)
Bankrupt Dummy                             (0.0428)
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:                            0.0628 (***)
Bankrupt+(HR x Bankrupt)                   (0.0238)
Adjusted [R.sup.2]           0.985          0.984
Observations              4350           4341
Panel B                    Property Taxes, Charges and Fees, and
                           Misc. Rev.
                             (1)           (2)
Bankrupt Count               0.0056        -0.0068
                            (0.0065)       (0.0160)
ACIR 1993 Home Rule x                       0.0156
Bankrupt Count                             (0.0150)
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:                            0.0088
Bankrupt+(HR x Bankrupt)                   (0.0059)
Adjusted [R.sup.2]           0.984          0.984
Observations              4350           4341

Panel A                    Property Taxes, Charges and Fees, and
                           Misc. Rev.
                            (3)              (4)
Bankrupt Dummy              -0.0686 (**)     -0.0613 (*)
                            (0.0346)         (0.0352)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home       0.1383 (***)
Rule 1 x Bankrupt Dummy     (0.0409)
Krane et al. 2002 Home                        0.1295 (***)
Rule 2 x Bankrupt Dummy                      (0.0424)
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:             0.0696 (***)     0.0683 (***)
Bankrupt+(HR x Bankrupt)    (0.0229)         (0.0232)
Adjusted [R.sup.2]           0.984            0.984
Observations              4341             4341
Panel B                   Property Taxes, Charges and Fees, and
                           Misc. Rev.
                             (3)             (4)
Bankrupt Count              -0.0232 (**)     -0.0236 (**0
                            (0.0117)         (0.0116)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home       0.0342 (***)
Rule 1 x Bankrupt Count     (0.0122)
Krane et al. 2002 Home                        0.0348 (***)
Rule 2 x Bankrupt Count                      (0.0122)
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:             0.0110 (*)       0.0112 (*)
Bankrupt+(HR x Bankrupt)    (0.0066)         (0.0067)
Adjusted [R.sup.2]           0.984            0.984
Observations              4341             4341

Panel A                     Property Taxes, Charges and Fees, and
                            Misc. Rev.
                              (5)
Bankrupt Dummy                0.0067
                             (0.0285)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x         0.0638
Bankrupt Dummy               (0.0419)
Combined Effect:              0.0705 (**)
Bankrupt+(HR x Bankrupt)     (0.0322)
Adjusted [R.sup.2]            0.983
Observations               2739
Panel B                    Property Taxes, Charges and Fees, and Misc.
                            Rev.
                            (5)
Bankrupt Count               -0.0044
                             (0.0108)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x         0.0178
Bankrupt Count               (0.0112)
Combined Effect:              0.0134
Bankrupt+(HR x Bankrupt)     (0.0085)
Adjusted [R.sup.2]            0.983
Observations               2739

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([PropTax.sub.it]) = [alpha]+ [[??] (Bankrupt Dummy * Post
*HomeRule).sup.r.sub.it] + [[beta](BankruptDummy *Post).sub.it]+
[[rho](HomeRule * Post).sup.r.sub.it] + [[theta](OperatingDummy *
Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [PropTax.sub.it] is revenue from property taxes, charges and
fees, and miscellaneous revenue in municipality i, in year t.
[BankruptDummy.sub.i] equals 1 when municipality i contains one or more
of the treatment chains. [Post.sub.t] is an indicator for whether or
not year t is after 2008 (the bankruptcy year). [HomeRule.sup.r.sub.i]
is a dummy variable equal to 1 if the municipality has home rule status
according to measure r, where r is one of the four home rule measures
discussed in section 3. OperatingDummy * Post controls for anny
operating stores after the bankruptcy year. [[delta].sub.i] represents
municipality fixed effects and [[gamma].sub.t] represents time fixed
effects. The "Combined Effect" row shows the sum of the coefficient on
BankruptDummy*Post and the coefficient on the interaction term
HomeRule*BankruptDummy * Post. This gives us the total effect of the
bankruptcy on home rule municipalities. Panel B of this table replaces
[BankruptDummy.sub.i] with [BankruptCount.sub.i], which is the total
number of bankrupt stores in municipality i. Standard errors clustered
at the county level are in parentheses. Municipalities are excluded if
they have over 50 big-box stores, have a 500% change between their
maximum total revenue (or sales tax or total expenditure) and their
minimum total revenue (or sales tax or total expenditure), or if they
are in the data for less than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.12: Post-Bankruptcy Comparison on Spending between Home Rule
and Non-Home Rule Cities

Panel A                          Total Expenditures
                            (1)            (2)

Bankrupt Dummy              -0.0228          -0.0495 (*)
                            (0.0170)         (0.0266)
ACIR 1993 Home Rule x                         0.0393
Bankrupt Dummy                               (0.0339)
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:                             -0.0102
Bankrupt+(HR x Bankrupt)                     (0.0217)
Adjusted [R.sup.2]           0.986            0.986
Observations              4350             4341
Panel B                          Total Expenditures
                            (1)            (2)
Bankrupt Count              -0.0170 (***)    -0.0274 (***)
                            (0.0065)         (0.0093)
ACIR 1993 Home Rule x                         0.0130
Bankrupt Count                               (0.0101)
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:                             -0.0145 (**)
Bankrupt+(HR x Bankrupt)                     (0.0072)
Adjusted [R.sup.2]           0.986            0.986
Observations              4350             4341

Panel A                       Total Expenditures
                            (3)              (4)

Bankrupt Dummy              -0.0441          -0.0458
                            (0.0342)         (0.0338)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home       0.0310
Rule 1 x Bankrupt Dummy     (0.0389)
Krane et al. 2002 Home                        0.0332
Rule 2 x Bankrupt Dummy                      (0.0385)
ICMA 1974 Home Rule x
Bankrupt Dummy
Combined Effect:            -0.0131          -0.0127
Bankrupt+(HR x Bankrupt)    (0.0194)         (0.0194)
Adjusted [R.sup.2]           0.986            0.986
Observations              4341             4341
Panel B                       Total Expenditures
                              (3)             (4)
Bankrupt Count              -0.0297 (***)    -0.0299 (***)
                            (0.0107)         (0.0106)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home       0.0155
Rule 1 x Bankrupt Count     (0.0114)
Krane et al. 2002 Home                        0.0157
Rule 2 x Bankrupt Count                      (0.0114)
ICMA 1974 Home Rule x
Bankrupt Count
Combined Effect:            -0.0142 (**)     -0.0142 (**)
Bankrupt+(HR x Bankrupt)    (0.0071)         (0.0071)
Adjusted [R.sup.2]           0.986            0.986
Observations              4341             4341

Panel A                       Total Expenditures
                               (5)

Bankrupt Dummy                -0.0379
                              (0.0280)
ACIR 1993 Home Rule x
Bankrupt Dummy
Krane et al. 2002 Home
Rule 1 x Bankrupt Dummy
Krane et al. 2002 Home
Rule 2 x Bankrupt Dummy
ICMA 1974 Home Rule x          0.0375
Bankrupt Dummy                (0.0384)
Combined Effect:              -0.0004
Bankrupt+(HR x Bankrupt)      (0.0250)
Adjusted [R.sup.2]             0.985
Observations                2739
Panel B                        Total Expenditures
                              (5)
Bankrupt Count                -0.0182 (*)
                              (0.0107)
ACIR 1993 Home Rule x
Bankrupt Count
Krane et al. 2002 Home
Rule 1 x Bankrupt Count
Krane et al. 2002 Home
Rule 2 x Bankrupt Count
ICMA 1974 Home Rule x          0.0012
Bankrupt Count                (0.0114)
Combined Effect:              -0.0170 (**)
Bankrupt+(HR x Bankrupt)      (0.0081)
Adjusted [R.sup.2]             0.985
Observations                2739

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([Expenditure.sub.it]) = [alpha]+ [[??](BankruptDummy * Post
*HomeRule).sup.r.sub.it] + [[beta](BankruptDummy * Post).sub.it] +
[[rho](HomeRule * Post).sup.r.sub.it] + [[theta](OperatingDummy *
Post).sub.it] + [[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [Expenditure.sub.it] is total expenditure in municipality i, in
year t. [BankruptDummy.sub.i] equals 1 when municipality i contains one
or more of the treatment chains. [Post.sub.t] is an indicator for
whether or not year t is after 2008 (the bankruptcy year).
[HomeRule.sup.r.sub.i] is a dummy variable equal to 1 if the
municipality has home rule status according to measure r, where r is
one of the four home rule measures discussed in section 3.
OperatingDummy*Post controls for any operating stores after the
bankruptcy year. [[delta].sub.i] represents municipality fixed effects
and [[gamma].sub.t] represents time fixed effects. The "Combined
Effect" row shows the sum of the coefficient on BankruptDummy * Post
and the coefficient on the interaction term HomeRule *BankruptDummy *
Post. This gives us the total effect of the bankruptcy on home rule
municipalities. Panel B of this table replaces [BankruptDummy.sub.i]
with [BankruptCount.sub.i], which is the total number of bankrupt
stores in municipality i. Standard errors clustered at the county level
are in parentheses. Municipalities are excluded if they have over 50
big-box stores, have a 500% change between their maximum total revenue
(or sales tax or total expenditure) and their minimum total revenue (or
sales tax or total expenditure), or if they are in the data for less
than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.13: Home Rule Results are Robust to Regional Controls

Panel A                           Own-Source Revenue
                                 (1)        (2)

Bankrupt Dummy               -0.1149 (***)    -0.1535 (***)
                             (0.0321)         (0.0439)
Krane et al. 2002 Home        0.1015 (***)     0.1203 (***)
Rule 1 x Bankrupt Dummy      (0.0354)         (0.0386)
Bankrupt+(HR x Bankrupt]     -0.0134          -0.0332
                             (0.0170)         (0.0277)
Controls x Bankrupt          -              CENSUS
Dummy                                       REGION
Adjusted [R.sup.2]           0.987             0.988
Observations              4341              4341
Panel B                          Own-Source Revenue
                                 (1)        (2)
Bankrupt Count               -0.0407 (***)    -0.0552 (***)
                             (0.0081)         (0.0158)
Krane et al. 2002 Home        0.0257 (***)     0.0301 (***)
Rule 1 x Bankrupt Count      (0.0089)         (0.0101)
                             -0.0150 (**)     -0.0251 (*)
+                            (0.0061)         (0.0136)
Controls x Bankrupt          -              CENSUS
Count                                       REGION
Adjusted [R.sup.2]           0.987             0.987
Observations              4341              4341

Panel A                           Own-Source Revenue
                              (3)                   (4)

Bankrupt Dummy                -0.0948 (**)       -0.1038 (***)
                              (0.0386)           (0.0276)
Krane et al. 2002 Home         0.0615 (**)        0.1460 (***)
Rule 1 x Bankrupt Dummy       (0.0312)           (0.0305)
Bankrupt+(HR x Bankrupt]      -0.0334             0.0422 (**)
                              (0.0282)           (0.0135)
Controls x Bankrupt         CENSUS             STATE
Dummy                       DIVISION
Adjusted [R.sup.2]             0.988              0.988
Observations                4341               4341
Panel B                          Own-Source Revenue
                              (3)              (4)
Bankrupt Count                -0.0501 (***)      -0.0483 (***)
                              (0.0158)           (0.0097)
Krane et al. 2002 Home         0.0248 (***)       0.0578 (***)
Rule 1 x Bankrupt Count       (0.0088)           (0.0103)
                              -0.0253 (*)         0.0095 (**)
+                             (0.0143)           (0.0043)
Controls x Bankrupt         CENSUS             STATE
Count                       DIVISION
Adjusted [R.sup.2]             0.988              0.988
Observations                4341               4341

Note: Panel A of this table reports estimates of regressions of the
following form:

ln([OwnSourceRev.sub.it]) = [alpha] + [[??] (BankruptDummy * Post
*HomeRule *RegionalDummy).sub.it] + [[beta](BankruptDummy * Post
*RegionalDummy).sub.it] + [[lambda](BankruptDummy *
RegionalDummy).sub.it] + [[rho](HomeRule*Post).sub.it] +
[[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

where [OwnSourceRev.sub.it] is own-source revenue in municipality i, in
year t. [BankruptDummy.sub.i] equals 1 when municipality i contains one
or more of the treatment chains. [Post.sub.t] is an indicator for
whether or not year t is after 2008 (the bankruptcy year).
[HomeRule.sub.i] is a dummy variable equal to 1 if the municipality has
home rule status according to Krane et al. (2002). OperatingDummy *
Post controls for any operating stores after the bankruptcy year.
BankruptDummyu*RegionalDummy controls for any bankrupt stores in
different census regions or states. [[delta].sub.i] represents
municipality fixed effects and [[gamma].sub.t] represents time fixed
effects. Bankrupt +(HR*Bankrupt) shows the sum of the coefficient on
BankruptDummy and the coefficient on the interaction term HomeRule *
BankruptDummy. Panel B of this table replaces [BankruptDummy.sub.i]
with [BankruptCount.sub.i], which is the total number of bankrupt
stores in municipality i. Standard errors clustered at the county level
are in parentheses. Municipalities are excluded if they have over 50
big-box stores, have a 500% change between their maximum total revenue
(or sales tax or total expenditure) and their minimum total revenue (or
sales tax or total expenditure), or if they are in the data for less
than 5 years.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.14: Home Rule Regression Discontinuity Analysis with Alternate
Bandwidths

                     (1)                           (2)
Panel A              First Stage, Municipality
                     has Home Rule

Population [greater   0.594 (***)                  0.572 (***)
than or equal to]
25,000
                     (0.169)                      (0.141)
Observations         79                          111
Panel B              Largest Percent Fall in
                     Revenue from 2010-2015
Municipality has     -0.044                       -0.088 (*)
Home Rule            (0.038)                      (0.045)
Observations         79                          111
Panel C              Percent Fall in Revenue
                     Greater Than 10%
Municipality has     -0.276                       -0.318 (**)
Home Rule            (0.177)                      (0.151)
Observations         79                          111
Panel D              Percent Fall in Revenue
                     Greater Than 30%
Municipality has     -0.060                       -0.161 (*)
Home Rule            (0.066)                      (0.085)
Observations         79                          111
Panel E              Municipality has Extremely
                     Strong Bond Rating
Municipality has      0.405                        0.360
Home Rule            (0.309)                      (0.276)
Observations        176                         234
Clusters             75                         106
Pop. Bandwidth       [+ or -]7,500               [+ or -]10,000

Note: Panel A of this table reports estimates of first-stage
regressions of the following form:

[HomeRule.sub.i] = [alpha] + [[beta](Above25000).sub.i] +
[[theta](Population).sub.i] + [[rho](Above25000 * Population).sub.i] +
[[epsilon].sub.i]

where [HomeRule.sub.i] is a dummy variable equal to 1 if municipality i
ever had home rule status between 2010 and 2015. [Above25000.sub.i] is
a dummy variable that equals to 1 if municipality's population exceeded
25,000 and equals to 0 otherwise. Population is the maximum number of
population municipality i had sometime between 1994 and 2009.
[Above25000.sub.*] Population is an interaction variable between
Above25000 and Population. This regression establishes a link between
the home rule population threshold in Illinois and a city's actual home
rule status. Panels B, C, D, and E show the results of regressions with
fuzzy regression discontinuity design, using instrumented HomeRule
variable to estimate various public-finance-related variables. Panel B
reports estimates of regressions of the following form:

[RevFall.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] + [[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall.sub.i] is the largest annual percentage fall in revenue
from 2010 to 2015 in municipality i. Panel C reports estimates of
regressions of the following form:

[RevFall10.sub.i] = [alpha] + [beta](HomeRule).sub.i] +
[[rho](Population).sub.i] + [[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall10.sub.i] is a dummy variable that equals to 1 if
municipality i has a fall in revenue larger than 10% at any point from
2010-2015. Panel D reports estimates of regressions of the following
form:

[RevFall30.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] +[[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [RevFall30.sub.i] is a dummy variable that equals to 1 if
municipality i has a fall in revenue larger than 30% at any point from
2010-2015. Panel E reports estimates of regressions of the following
form:

[StrongBond.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[rho](Population).sub.i] +[[lambda](HomeRule * Population).sub.i] +
[[epsilon].sub.i]

where [StrongBond.sub.i] is a dummy variable that equals to 1 if
municipality i has extremely strong bond rating in IL data (from 1994
to 1996) or in scraped data (2015).

Regressions from all five panels are run with two different population
bandwidths. Column 1 includes cities with populations between 17,500
and 32,500, and column 2 includes cities with populations between
15,000 and 35,000. Standard errors clustered at the municipality level
are in parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.15: Polynomial Robustness Checks for Home Rule Regression
Discontinuity Analysis

                    (1)             (2)
Panel A              First Stage, Municipality has Home Rule

Population[greater    0.575 (***)     0.588 (***)
than or equal
to]25,000
                     (0.188)         (0.168)
Observations        148             183
Panel B             Largest Percent Fall in Revenue from 2010-2015
Municipality has     -0.038          -0.089 (*)
Home Rule            (0.042)         (0.047)
Observations        148             183
Panel C             Percent Fall in Revenue Greater Than 10%
Municipality has     -0.319          -0.356 (*)
Home Rule            (0.212)         (0.183)
Observations        148             183
Panel D              Percent Fall in Revenue Greater Than 30%
Municipality has     -0.039          -0.141 (*)
Home Rule            (0.071)         (0.085)
Observations        148             183
Panel E             Municipality has Extremely Strong Bond Rating
Municipality has      0.419           0.366
Home Rule            (0.327)         (0.313)
Observations        283             331
Clusters            135             162
Pop. Bandwidth      [+ or -]12,500  [+ or -]15,000

                        (3)           (4)
Panel A               First Stage, Municipality has Home Rule

Population[greater      0.605 (***)     0.573 (***)
than or equal
to]25,000
                       (0.148)         (0.140)
Observations          259             314
Panel B               Largest Percent Fall in Revenue from 2010-2015
Municipality has       -0.089 (*)      -0.085
Home Rule              (0.048)         (0.053)
Observations          257             312
Panel C               Percent Fall in Revenue Greater Than 10%
Municipality has       -0.331 (**)     -0.332 (**)
Home Rule              (0.159)         (0.161)
Observations          258             313
Panel D               Percent Fall in Revenue Greater Than 30%
Municipality has       -0.160 (*)      -0.160
Home Rule              (0.092)         (0.102)
Observations          258             313
Panel E               Municipality has Extremely Strong Bond Rating
Municipality has        0.325           0.344
Home Rule              (0.289)         (0.293)
Observations          434             498
Clusters              218             257
Pop. Bandwidth        [+ or -]18,000  [+ or -]20,000

Note: Panel A of this table reports estimates of first-stage
regressions of the following form:

[HomeRule.sub.i] = [alpha] + [[beta](Above25000).sub.i] +
[[theta](Population).sub.i] + [[delta]([Population.sup.2]).sub.i] +
[[gamma](Above25000 * Population).sub.i] + [[pi](Above25000 *
[Population.sup.2]).sub.i] + [[epsilon].sub.i]

where [HomeRule.sub.i] is a dummy variable equal to 1 if municipality i
ever had home rule status between 2010 and 2015. [Above25000.sub.i] is
a dummy variable that equals to 1 if municipality's population exceeded
25,000 and equals to 0 otherwise. Population is the maximum number of
population municipality i had sometime between 1994 and 2009.
Above25000 * Population is an interaction variable between Above25000
and Population. This regression establishes a link between the home
rule population threshold in Illinois and a city's actual home rule
status. Panels B, C, D, and E show the results of regressions with
fuzzy regression discontinuity design, using instrumented HomeRule
variable to estimate various public-finance-related variables. Panel B
reports estimates of regressions of the following form:

[RevFall.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[theta](Population).sub.i] + [[delta]([Population.sup.2]).sub.i] +
[[gamma](HomeRule * Population).sub.i] + [[pi](HomeRule *
[Population.sup.2]).sub.i] + [[epsilon].sub.i]

where [RevFall.sub.i] is the largest annual percentage fall in revenue
from 2010 to 2015 in municipality i. Panel C reports estimates of
regressions of the following form:

[RevFall10.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[theta](Population).sub.i] + [[delta]([Population.sup.2]).sub.i] +
[[gamma](HomeRule * Population).sub.i] + [[pi](HomeRule *
[Population.sup.2]).sub.i] + [[epsilon].sub.i]

where [RevFall10.sub.i] is a dummy variable that equals to 1 if
municipality i has a fall in revenue larger than 10% at any point from
2010-2015. Panel D reports estimates of regressions of the following
form:

[RevFall30.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[theta](Population).sub.i] + [[delta]([Population.sup.2]).sub.i] +
[[gamma](HomeRule * Population).sub.i] + [[pi](HomeRule *
[Population.sup.2]).sub.i] + [[epsilon].sub.i]

where [RevFall30.sub.i] is a dummy variable that equals to 1 if
municipality i has a fall in revenue larger than 30% at any point from
2010-2015. Panel E reports estimates of regressions of the following
form:

[StrongBond.sub.i] = [alpha] + [[beta](HomeRule).sub.i] +
[[theta](Population).sub.i] + [[delta]([Population.sup.2]).sub.i] +
[[gamma](HomeRule * Population).sub.i] + [[pi](HomeRule *
[Population.sup.2]).sub.i] + [[epsilon].sub.i]

where [StrongBond.sub.i] is a dummy variable that equals to 1 if
municipality i has extremely strong bond rating in IL data (from 1994
to 1996) or in scraped data (2015). Regressions from all five panels
are run with four different population bandwidths. Column 1 includes
cities with populations between 12,500 and 37,500; column 2 includes
cities with populations between 10,000 and 40,000; column 3 includes
cities with populations between 7,000 and 43,000; and column 4 includes
cities with populations between 5,000 and 45,000. Standard errors
clustered at the municipality level are in parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.16: Sales Tax and Gross Receipts Robustness Checks for
Different Operating Store Controls

Panel A                 Sales Tax and Gross Receipts
                        (1)          (2)

Bankrupt Dummy        -0.1575 (***)    -0.1494 (***)
                      (0.0383)         (0.0421)
Constant               9.8201 (***)     9.8206 (***)
                      (0.0254)         (0.0252)
Operating Control   Operating Dummy  Operating Count x
                    x Post           Post
Adjusted [R.sup.2]     0.947            0.947
Observations        4346             4346
Panel B               Sales Tax and Gross Receipts
                    (1)              (2)
Bankrupt Count        -0.0307 (***)    -0.0418 (***)
                      (0.0088)         (0.0118)
Constant               9.8210 (***)     9.8212 (***)
                      (0.0260)         (0.0257)
Operating Control   Operating Dummy  Operating Count x
                    x Post           Post
Adjusted [R.sup.2]     0.947            0.947
Observations        4346             4346

Panel A                Sales Tax and Gross Receipts
                       (3)           (4)

Bankrupt Dummy        -0.1465 (***)    -0.1467 (***)
                      (0.0456)         (0.0386)
Constant               9.8205 (***)     9.8205 (***)
                      (0.0253)         (0.0252)
Operating Control   Big Box Count x  No Control
                    Post
Adjusted [R.sup.2]     0.947            0.947
Observations        4346             4346
Panel B               Sales Tax and Gross Receipts
                    (3)              (4)
Bankrupt Count        -0.0552 (***)    -0.0308 (***)
                      (0.0196)         (0.0087)
Constant               9.8212 (***)     9.8211 (***)
                      (0.0257)         (0.0259)
Operating Control   Big Box Count x  No Control
                    Post
Adjusted [R.sup.2]     0.947            0.946
Observations        4346             4346

Note: This table reports estimates of regressions of the following
form:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

where [Revenue.sub.it] is sales tax and gross receipt revenue for
municipality i, in year t; [BankruptDummy.sub.i] in Panel A equals 1
when municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not the year is after 2008,
the bankruptcy year. Operating Dummy * Post controls for whether or not
municipality i contains any operating stores in the treatment category
after the bankruptcy year. In Column 2 OperatingDummy is replaced by
OperatingCount. Similarly in Columns 3 and 4 BigBoxCount and no
controls are used instead of OperatingDummy, respectively. Panel B
follows the below specification:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptCount * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

Where [BankruptCount.sub.i] is substituted for [BankruptDummy.sub.i] in
Panel A and equals the number of bankruptcies of treatment chains in
municipality i. In both panels, [[delta].sub.i] represents municipality
fixed effects and [[gamma].sub.t] represents year fixed effects.
Similarly to panels A, Columns 1-4 show different operating
specifications and no control. Standard errors are in parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.17: Own-Source Revenue Robustness Checks for Different
Operating Store Controls

Panel A                              Own-Source Revenue
                       (1)           (2)

Bankrupt Dummy        -0.0499 (***)    -0.0368 (**)
                      (0.0142)         (0.0160)
Constant              11.3665 (***)    11.3665 (***)
                      (0.0071)         (0.0071)
Operating Control   Operating Dummy  Operating Count x
                    x Post           Post
Adjusted [R.sup.2]     0.988            0.988
Observations        4350             4350
Panel B                              Own-Source Revenue
                    (1)              (2)
Bankrupt Count        -0.0182 (***)    -0.0191 (***)
                      (0.0040)         (0.0058)
Constant              11.3663 (***)    11.3664 (***)
                      (0.0070)         (0.0070)
Operating Control   Operating Dummy  Operating Count x
                    x Post           Post
Adjusted [R.sup.2]     0.988            0.988
Observations        4350             4350

Panel A                              Own-Source Revenue
                        (3)             (4)

Bankrupt Dummy         -0.0247          -0.0447 (***)
                       (0.0177)         (0.0145)
Constant               11.3663 (***)    11.3667 (***)
                       (0.0071)         (0.0071)
Operating Control    Big Box Count x  No Control
                     Post
Adjusted [R.sup.2]      0.988            0.988
Observations         4350             4350
Panel B                              Own-Source Revenue
                     (3)              (4)
Bankrupt Count         -0.0203 (**)     -0.0182 (***)
                       (0.0096)         (0.0040)
Constant               11.3664 (***)    11.3663 (***)
                       (0.0070)         (0.0070)
Operating Control    Big Box Count x  No Control
                     Post
Adjusted [R.sup.2]      0.988            0.988
Observations         4350             4350

Note: This table reports estimates of regressions of the following
form:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptDummy * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

where [Revenue.sub.it] is own-source revenue for municipality i, in
year t; [BankruptDummy.sub.i] in Panels A equals 1 when municipality i
contains one or more of the treatment chains. [Post.sub.t] is an
indicator for whether or not the year is after 2008, the bankruptcy
year. OperatingDummy * Post controls for whether or not municipality i
contains any operating stores in the treatment category after the
bankruptcy year. In Column 2 OperatingDummy is replaced by
OperatingCount. Similarly in Columns 3 and 4 BigBoxCount and no
controls are used instead of OperatingDummy, respectively. Panels B
follows the below specification:

ln([Revenue.sub.it]) = [alpha] + [[beta](BankruptCount * Post).sub.it]
+ [[theta](OperatingDummy * Post).sub.it] + [[delta].sub.i] +
[[gamma].sub.t] + [[epsilon].sub.it]

Where [BankruptCount.sub.i] is substituted for [BankruptDummy.sub.i] in
Panels A and equals the number of bankruptcies of treatment chains in
municipality i. In both panels, 5i represents municipality fixed
effects and [[gamma].sub.t] represents year fixed effects. Similarly to
panels A, Columns 1-4 show different operating specifications and no
control. Standard errors are in parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01

Table A.18: Total Expenditure Robustness Checks for Different Operating
Store Controls

Panel A                     Total Expenditure
                        (1)             (2)

Bankrupt Dummy         -0.0336 (**)       -0.0228
                       (0.0164)           (0.0170)
Constant               11.5219 (***)      11.5220 (***)
                       (0.0091)           (0.0091)
Operating Control      Operating Dummy  Operating Count x
                       x Post           Post
Adjusted [R.sup.2]     0.986               0.986
Observations        4350                4350
Panel B                 Total Expenditure
                     (1)                (2)
Bankrupt Count         -0.0145 (***)      -0.0170 (***)
                       (0.0040)           (0.0065)
Constant               11.5216 (***)      11.5218 (***)
                       (0.0091)           (0.0091)
Operating Control    Operating Dummy    Operating Count x
                     x Post             Post
Adjusted [R.sup.2]      0.986              0.986
Observations         4350               4350

Panel A             Total Expenditure
                      (3)                 (4)

Bankrupt Dummy        -0.0119            -0.0282 (*)
                      (0.0176)           (0.0164)
Constant              11.5218 (***)      11.5221 (***)
                      (0.0091)           (0.0091)
Operating Control   Big Box Count x    No Control
                    Post
Adjusted [R.sup.2]     0.986              0.986
Observations        4350               4350
Panel B             Total Expenditure
                    (3)                 (4)
Bankrupt Count        -0.0201 (*)        -0.0200 (*)
                      (0.0111)           (0.0111)
Constant              11.5218 (***)      11.8138 (***)
                      (0.0091)           (0.0086)
Operating Control   Big Box Count x    No Control
                    Post
Adjusted [R.sup.2]     0.986              0.986
Observations        4350               4350

Note: This table reports estimates of regressions of the following
form:

ln([Expenditure.sub.it]) = [alpha] + [[beta](BankruptDummy *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] +
[[delta].sub.i] + [[gamma].sub.t] + [[epsilon].sub.it]

where [Expenditure.sub.it] is the total expenditures for municipality
i, in year t; [BankruptDummy.sub.i] in Panels A equals 1 when
municipality i contains one or more of the treatment chains.
[Post.sub.t] is an indicator for whether or not the year is after 2008,
the bankruptcy year. Operating Dummy * Post controls for whether or not
municipality i contains any operating stores in the treatment category
after the bankruptcy year. In Column 2 OperatingDummy is replaced by
OperatingCount. Similarly in Columns 3 and 4 BigBoxCount and no
controls are used instead of OperatingDummy, respectively. Panels B
follows the below specification:

ln([Expenditure.sub.it]) = [alpha] + [[beta](BankruptCount *
Post).sub.it] + [[theta](OperatingDummy * Post).sub.it] +
[[delta].sub.i][[gamma].sub.t] + [[epsilon].sub.it]

Where [BankruptCount.sub.i] is substituted for [BankruptDummy.sub.i] in
Panels A and equals the number of bankruptcies of treatment chains in
municipality i. In both panels, [[delta].sub.i] represents municipality
fixed effects and [[gamma].sub.t] represents year fixed effects.
Similarly to panels A, Columns 1-4 show different operating
specifications and no control. [[delta].sub.i] represents municipality
fixed effects and [[gamma].sub.t] represents time fixed effects.
Standard errors are in parentheses.

(*) p < 0.1, (**) p < 0.05, (***) p < 0.01
COPYRIGHT 2019 The American Enterprise Institute
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Shoag, Daniel; Tuttle, Cody; Veuger, Stan
Publication:AEI Paper & Studies
Date:Jul 1, 2019
Words:23264
Previous Article:China's Global Business Footprint Shrinks.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters