Printer Friendly

Assessor incentives and property assessment.

1. Introduction

The taxation of property is one of the most important sources of tax revenue in local public finance. According to the U.S. Census of Governments, local governments collected 45.2% of their tax revenue from the taxation of their residents' property in the 2005 fiscal year. This is a much larger share than any other form of taxation, and it is critical to the provision of local public goods such as education. Yet it is also widely accepted that the property tax is among the most disliked of taxes. (1) This is particularly true during periods of rising real estate prices when the level of taxation does not necessarily correspond with the income growth of the homeowners, leading them to express feelings of "being taxed out of their homes." Property tax assessors and their methodologies often become the target of this frustration, causing them to become the focus of taxpayer pressure to reduce the burden. (2) Furthermore, depending on the jurisdiction, property assessors may be elected directly by the local constituency or appointed to this office by another elected official. This suggests there is an incentive for incumbent officials facing reelection to underassess property values to lower the effective tax burden of their constituency in a vote-seeking effort.

If assessors are not appraising property at their true market value and instead are appealing to a political constituency, there are a variety of implications. In terms of equity, assessors may discriminate between household or property types for underassessment and alter the incidence of the property tax burden along a dimension of political power. This would be particularly true if assessors compensate the underassessment of one household type with the overassessment of another, which they would have an incentive to do if they are required to comply with the International Association of Assessing Officer (IAAO) standards or other state rules of a similar nature. (3) For instance, Black (1977) has found that effective property taxes in 1960 Boston were more regressive than were previously thought after it was taken into account that assessors tended to understate the value of property at the higher end of the market.

While property taxes were handled almost exclusively at the local levels of government historically, state involvement in property tax administration has become increasingly important over the past few decades. This has occurred in part as a result of the wave of school finance centralization reforms that took place throughout the 1980s and 1990s. These reforms came via a variety of state supreme court rulings and state legislative action on per pupil education spending inequalities among school districts (Fischel 200la). (4) In the 2005 fiscal year, the Census of Governments reported that state property tax revenues averaged about 1.7% of all own-source state tax revenues, with Vermont deriving the largest share at 35.3%. In 2007, all but 12 states collected a share of property tax revenue, either by levying their own tax rate or by collecting a portion of the local governments' property tax revenues. (5) In addition to serving as a revenue source, several states use the level of assessed property values as a measure of ability-to-pay in their school district aid formulas of state funding. For example, West Virginia's school-aid formula increases the level of state funding to the school district if they have a decrease in the value of their assessed property. So in West Virginia, which strictly uses elected assessors, if the assessor lowers the appraisals they not only cull favor from the constituency for reducing their individual property tax burden, but they also increase their share of state funding. Furthermore, underassessment creates distortions in the voters' choices when ordering preferences on statewide property tax legislation if they do not bear the full burden of the bill. Finally, the federal deductability of property taxes may potentially result in additional exporting of the property tax burden across states. What is clear is that the ability to measure property at market value accurately, not just uniformly, is becoming increasingly important.

While the elected assessor is both bureaucrat and politician, even if the assessor is appointed it is possible that their elected appointer will apply pressure to underassess constituent property. In The Role of the States in Strengthening the Property Tax, the Advisory Commission on Intergovernmental Relations (ACIR 1963) dismissed the issue of appointed versus elected assessors on the point that they may be susceptible to political pressure. Though indirect, it could be true that the voters would just take their complaints to the elected official that appoints the assessor, who then could pressure the assessor to lower property assessments. Alternatively, one could see a situation where appointed assessors serve as a convenient scapegoat for elected officials to raise tax revenue with higher assessments rather than increasing the property tax rate.

The purpose of this article is to test for sources of political pressure as well as differences in the levels of assessment between appointed and elected property tax assessors. This is tested using the median assessment-to-sale price ratio (sales ratio) for Virginia counties and independent cities, calculated by the Virginia Department of Taxation (VDT 2007), as the dependent variable. Virginia is one of the few states that allows its local governments to have either appointed or elected assessors, which permits a direct comparison of the types on the level of the sales ratio. The results indicate that elected assessors underassess more than appointed assessors by 1 to 2 standard deviations (SDs) of the sales ratio. This is after controlling for a variety of constituent socioeconomic and district characteristics that themselves influence the sales ratio in a manner consistent with what would be expected if assessors are responsive to factors that might promote their political support.

2. Previous Research

Political economy issues for local governments have by no means received a lack of attention from economists, especially concerning the role of local property taxes. Fischel (2001b) supplied what he termed a "homevoter model," where zoning combined with local property taxes led to the more efficient provision of local public goods when compared with state level provision. Since present property values quickly capitalize the expected future value of amenities such as local public goods, Glaeser (1995) argues that even myopic politicians have the incentive to consider long-run implications of public goods that are funded by property taxes. (6) What is interesting in the role of the assessor here is the ability of political pressure to undermine these incentives. While a local voter may vote for community-level property taxes set to some ideal level that corresponds with their demand for local public goods, their individual incentive to minimize personal tax burden can be achieved by pressuring the local assessor's office.

To illustrate these conflicting tensions, Johnson (1989) provides a behavioral model where the assessor maximizes wealth from office by gaining political support subject to those regulatory constraints. This is modeled differently for appointed and elected assessors because they are concerned with different constituent bases. In the case of the assessor appointed by someone in a different political office, they maximize political support by serving as scapegoats and assessing at higher values so that their appointers raise revenues without the political fallout of higher taxes. The elected assessors, on the other hand, maximize political support by providing as much tax relief as possible by lowering assessments among the parcels. In either case, the presence of commercial property allows for assessors to export the tax and allows them to further raise revenues and comply with regulatory constraints surrounding the level of assessment.

Interestingly, the Johnson (1989) model makes this assumption regarding differential treatment of elected and appointed assessors in spite of a long standing disregard of the issue. The ACIR (1963), whose report seems to have driven much of the research over the decades that followed, had disregarded any differences due to political pressure on the grounds that constituents could just as easily pressure the assessor's appointer(s) for lower assessments. This view holds some empirical support in the previous literature. There has been considerable attention paid in the recent literature on the policy restraints on rising local property taxes during the rapid acceleration of housing prices around the United States at the turn of the 21st century (Bowman 2006; Cornia and Walters 2006). Since these policy changes are prescribed by lawmakers rather than assessors, there is likely some credence to the idea that appointers can request their assessors to lower the level of assessment. Since both conflicting propositions regarding the incentives of appointed assessors have intuitive appeal, their susceptibility to political pressure from the local constituency to underassess is primarily an empirical question.

When differences in the assessor type have risen in the empirical literature they have almost entirely focused on the assessor type's influence on horizontal equity via uniformity measures such as the coefficient of dispersion (COD). Lowery (1984) found that uniformity erodes under fiscal stress and tax limitations when the assessor is elected but strengthens under appointed assessors. Strauss and Sullivan (1998) tested the influence of several indicators of assessor authority, state requirements, the level of government responsible for assessment, and assessor type on the level of uniformity. They found that elected assessors had higher levels of uniformity, but that its effect diminished as the office moved away from being elected at the local level toward the county or state level. Bowman and Mikesell (1989), in testing for determinants of real property assessment uniformity in Virginia, found the inclusion of an indicator variable for assessor type to be insignificant. Historically this focus on uniformity makes a great deal of sense because the property tax was a local government affair. Since the property tax rate was determined as the total municipal levies divided by the total assessed values, underassessment would not be a problem if the underassessment was uniform across households. However, with the state collecting and redistributing property tax revenues, underassessment does become problematic even if they continue to determine their property tax rate in the same manner, which often they do not.

Unfortunately the COD implies nothing about the proclivity of an assessor to engage in any form of systematic bias (Stewart 1977). This is due to the fact that the COD serves a measure of dispersion around the localities' own median sales ratio but does not indicate whether that median ratio is at the required level (Payton 2006). In order to directly test for characteristics that influence the degree of underassessment, the sales ratio is the most relevant available measure. Literature using this statistic as the dependent variable is almost nonexistent. Footnote 9 of Bowman and Mikesell (1989) mentions that an attempt to use the sales ratio instead of the COD carried an R2 of just 0.04 with no significance in any explanatory variables, including assessor type. Yet the intention of that work was to look for factors influencing the dispersion of the sales ratio, not the level of the sales ratio itself. As a result, the regressors consist of factors that might affect the ability of an assessor to accurately assess property, such as whether the assessor was full or part time, level of assessor certification, and availability of assessment maps. While we would expect these variables to affect the variance of a measure, it should not affect the expected value of the sales ratio absent any systematic bias. In other words, explanatory variables controlling for the level of difficulty do not explain why the degree of error always lands on the side of being below true market value.

Lowery (1982) is the only previous article I am aware of that has used the sales ratio as the dependent variable. The regressions are based on a mail survey of Michigan assessors and employ ordinary least squares (OLS) on a variety of variables that primarily control for factors that proxy for the difficulty of assessment, as well as an indicator for whether the assessor was appointed. The overall work is insightful and very good, but the shortcoming of the article is that the econometric model starts from the perspective that underassessment is the result of poor information. The author then starts searching for variables indicating information quality that might lead to higher sales ratios (i.e., certification, availability of tax maps, computers, etc.). As in Bowman and Mikesell (1989), these are factors that would influence the dispersion of sales ratios but not cause a systematic bias. Instead of variables that would improve information, there should be variables that constrain or bias the assessor away from maintaining the mandated sales ratio. This is the approach taken in this article.

One additional area of research this work contributes to is in the literature specifically focused on the issue of appointed versus elected public officials. Alesina and Tabellini (2008) defined a multiple task model of appointed and elected officials to determine the socially optimal allocation of tasks. Ultimately they conclude that appointed bureaucrats should be allocated tasks that require greater technical skill and long-term consistency, while elected officials are better suited to tasks dependent on uncertain social preferences or whether compensation schemes are needed for Pareto improvements. Empirically, Besley and Coate (2003) found elected utility regulators to be considerably more proconsumer in the United States than their appointed counterparts. Hanssen (1999) found that judicial rulings were less predictable in state courts where judges are appointed, suggesting greater autonomy from political pressure. Finally, Hoover (2008) found no statistically significant effect on student performance in Alabama school districts when either the school board or superintendent were elected.

3. Background on Virginia

In testing for sources of political pressure on assessors, comparisons across states would be problematic, since there is considerable heterogeneity in the level of accountability, certification standards, and legal authority of the assessors (Strauss and Sullivan 1998). (7) For this reason, studying assessors in a state with a mixed system of both appointed and elected assessors becomes appropriate. Very few states have a truly mixed system of assessor types, and even those that do allow for it usually employ one type overwhelmingly or otherwise determine it endogenously. For instance, New York has a mixed system but allows for elected assessors only in the cases of villages and other small municipalities. Virginia has several attractive characteristics for determining the factors that influence the sales ratio. Virginia's constitution allows for a district to choose its own assessor type, and as a result 48 of its 134 districts have appointed assessors, while the remaining districts leave the responsibility of maintaining assessment records to elected officials, typically the Commissioner of Revenue. (8) While few elected officials actually perform the assessments themselves, instead opting to employ private firms, they are responsible for the provision of final assessments as well as other important duties pertaining to the tax rolls and land books. These are the officials that property owners wishing to appeal their assessment are encouraged or required to speak with first. Interestingly, Virginia has "Truth in Taxation Laws" that require the constituency to be told that the assessors are not responsible for their tax burden, but just the assessment of their property. These laws seem to be in and of themselves recognition of constituent pressure on assessors.


Virginia is unique in that the type of assessor chosen by a district is written and formalized into the individual district's charter, often many decades ago. I can find no historical document that catalogs a debate over the type of assessor for any district. They all serve four year renewable terms, with the most recent election being in 2007, though assessment data are not yet available for that year. There is also precedence for using Virginia to estimate differences between appointed and elected assessors, with Bowman and Mikesell (1989) using it to estimate the influence of assessor type on the COD. Figure 1 illustrates the assessor type by district in Virginia.

A district's property tax rate is one of the policies determined by a committee, usually known as the County Board of Supervisors, the members of which are elected to office. The local property tax rate is frequently a subject of contention among policy makers as well as the electorate. During the period the data were collected, the appreciation of housing prices resulted in budget surpluses and generated numerous editorials, letters to the editor, and coverage of debates in the local newspapers over how these surpluses should be used (e.g., lowering property tax rates, increased spending on services, etc.).

The median sales ratio for each of the 134 districts is reported in a study by the VDT from 2001 to 2006. The study sampled arms' length market sales of existing residential and commercial properties that took place over the year then divided it by the assessed value of the property the locality had on the books at the time of its sale. For each of these jurisdictions, the median of these sales ratios is reported in the study. A sales ratio less than 1.0 would suggest the property was underassessed, while a value above 1.0 would suggest overassessment. Virginia's tax code stipulates that assessors must demonstrate accurate assessment with sales ratios above 0.70 or else the state may withhold the locality's share of the state profits from the sale of alcoholic beverages. (9)

Furthermore, the state uses the median sales ratio in its school-aid formula, so that districts that underassess receive a smaller share of the state funding for education.


Figure 2 is a histogram of the median sales ratios in the years of study in this article, 2001 to 2006. It can easily be seen in Figure 2 that property underassessment was the norm.

Over the period of the data, the average of the median sales ratios for elected assessors was 0.85 versus 0.84 for appointed. Both the highest and the lowest median sales ratio came from a district with an appointed assessor. In fact, the three observations that appear as outliers on the minimum side of the histogram in Figure 2 come from a district with an appointed assessor.

The observations used in the regression analysis are the districts in the years they conduct a reassessment of properties. Only 55 of the 134 districts conduct annual reassessments, and therefore they are the only districts to appear on an annual basis in the dataset. Virginia's tax code stipulates that reassessments must be conducted at least every four years, with exceptions for very small districts that may have up to six years. One of the caveats of the design of the sales ratio studies is that properties that undergo reassessment early in the year might appear undervalued in the sales ratio study if it is sold later in the same year. This may lead to misleadingly low sales ratios in areas with greater housing price appreciation, such those near Washington, D.C., in northem Virginia. Since housing prices are not available during this period for all counties, to control for such growth effects this article follows Bowman and Mikesell (1989) and includes the average annual population growth rate as a proxy variable.

The state of Virginia sets the election cycle such that all districts do not have their elections in the same year. Instead, districts are assigned to one of two timelines so that there are district elections across the state every two years, while each individual district only has an election every four years. In this data set, the years 2001, 2003, and 2005 were election years. Figure 3 reports the average median sales ratio of the districts that carried out assessments in each year and provides a breakdown according to whether or not they had an election. No clear pattern of an election cycle underassessment appears in Figure 3. In fact, counties reassessing during an election year actually had higher sales ratios on average than counties that were not. The average sales ratio during election years also does not differ much between elected and appointed assessors. In election years, the average sales ratio for an appointed assessor is 0.851, and that for an elected assessor is 0.842.


4. Model Specification

Like Johnson (1989), the empirical model is motivated by considering the assessor's objective to be political support maximizing. Constituent voters who are unhappy with their assessment take their complaints either to the assessor directly or to the assessor's appointer. In Virginia, districts that require appointed assessors are usually chosen by the Commissioner of Revenue, another elected representative, or in a handful of cases they are appointed by the County Board of Supervisors, which also consists of elected representatives. Trying to appease voters provides an incentive to underassess, but assessors maximizing political support have a counterincentive to increase the assessment rates to comply with state law. Elected assessors often try to garner endorsements from other elected officials, and any assessor would likely find themselves in an unpopular position with other elected officials if low assessments created fiscal stress or required an increase in tax rates. As discussed in the previous section, assessors conceivably could even be pressured by elected officials to increase their assessments in lieu of increasing tax rates. Finally, because of statewide property tax sharing formulas, the state has constraints on minimum sales ratios that were discussed in the previous section. Therefore, an assessor who may underassess one group might try to inflate their assessment rate by increasing the rate of assessment on another.

Since these conflicting incentives would arise out of different sources of political pressure, the econometric model will test various social and economic characteristics of the districts in an attempt to estimate the significance of these possible factors. Letting Rit represent the median sales ratio for Virginia district i undergoing reassessment in year t, the specification of the model will try to explain variation in the sales ratio using voter population characteristics (income, racial homogeneity, share of population over age 65), district characteristics (election year, population growth rate, commercial property base, property tax rate, fiscal stress), frequency of reassessments (Freq), and an indicator for whether or not the assessor is elected:

[] = [[beta].sub.0] + [[beta].sub.1] [Elected.sub.i] + [[beta].sub.2] [] + [[beta].sub.3] [] + [[beta].sub.4]Freq + [[epsilon]]. (4.1)

The definition and sources for these variables are listed in Table 1, while their descriptive statistics are displayed in Table 2. The rest of this section will describe the expectations behind these variables that might warrant their inclusion.

One of the most common independent variables in the previous literature on assessor behavior is the proportion of the population that is African American. To some degree this dates the existing literature, since racial homogeneity/fractionalization has increasingly become the control for minority effects in political economy issues. (10) Since there are several Virginia counties where African Americans are in the majority, the more contemporary literature is followed in this article with the use of a racial homogeneity score, calculated in the manner of the Herfindahl-Hirschman Index where larger values indicate a greater prominence of a single race. If Virginia assessors want to appease the state comptroller with higher median assessments while trying to please the median voter with lower valuations, then assessors might target minorities for higher assessments, in which case the expected sign on racial homogeneity would be positive. To illustrate this intuition, consider two districts that differ in their racial homogeneity but are otherwise identical. Suppose one county has an equal split of four races while the other county has a population that is 99% white, giving the former county a comparatively low racial homogeneity score. An assessor in the racially homogeneous county using race as a signal for preferences would likely see whites as the median voter to be underassessed, allowing them to compensate that lower assessment with higher assessments for the minority groups occupying the remaining 1%. Whereas in the diverse county, no race would be any more likely to contain the median voter than any other, which would restrain the assessor from targeting any specific group for overassessment.

Senior citizens, those over the age of 65, might also be prime candidates for underassessment for several reasons. First, they are more likely to own their home and have paid off their mortgage leaving the property tax as their primary "rent" expense, and as a result they may be more sensitive to changes in their assessment. Second, they are less likely to be earning income and thus do not reap any benefits from being able to deduct property taxes. Finally, the political conventional wisdom considers them to be more prevalent and informed voters that are likely to turn out on election day. Thus the intuition would be that the higher the proportion of senior citizens in the population, the more pressure the assessor would face, contributing to a lower sales ratio.

The traditional view of high-income households in the property tax assessment literature has been that voters with more income are better informed and educated about their property taxes and would likely pressure their assessors for more accurate assessments.

While this sits well with the uniformity literature, it does not obviously translate into an argument for lower sales ratios because of the tax treatment of housing across income groups. Higher income groups are more likely to itemize and accept the deduction of property taxes from their taxable income, whereas lower income groups are more apt to take the standard deduction. Itemization means that a higher assessment is more fully realized in the lower income groups than those of the higher because of its direct effect in the property tax bill. Since there are two countering income effects, the sign of income is ambiguous.

If assessors see commercial property as a means of exporting the property tax burden to nonresident owners, then they might overassess that property relative to resident-owned property (Johnson 1989). Intuitively, since commercial property owners are more likely to be nonresidents and therefore nonvoters, they would be favorable targets for overassessment. The ratio study done by the VDT provides a breakdown of the number of sales included in the study that were commercial property for each district. This is included as a share of the total number of sales for the district, with the expectation that it will be positively correlated with the sales ratio.

If property owners are more sensitive to their assessments when they face a high property tax rate, then it should have a negative correlation with the sales ratio (Bowman and Mikesell 1989). One complication with the property tax rate could arise if the policy makers determining the property tax rate have the ability to pressure the assessors. Intuitively, policy makers could potentially lower the nominal property tax rates but offset their revenue consequences with higher property assessments. The result would be multicollinearity between the property tax rate and the assessor type. However, neither variance inflation factors nor a Belsley, Kuh, and Welsch (1980) variance-decomposition proportion matrix were indicative of such multicollinearity. Furthermore, on average districts with appointed assessors have slightly higher property tax rates by about one-quarter of a point. Nevertheless, three districts counting for a total of 15 observations in the panel are excluded because the governing body setting the property tax rate plays a role in appointing the assessors in those areas. (11) While the presented results will have excluded these observations, including them makes no qualitative difference. (12)

Additionally, Lowery (1984) has previously found that fiscal stress has had eroding effects on uniformity in assessments. The district's unemployment rate, as well as a composite measure of district government fiscal stress provided by the Virginia Department of Housing and Community Development (VDHCD), is included to capture a similar effect in the sales ratio. (13) The expectation is that higher unemployment and fiscal stress could result in higher sales ratios and therefore have positive signs on the coefficients. Conflictingly, unemployed residents may be more prone to pressure officials for lower assessments, so for unemployment the expected sign is indeterminate.

Equation 4.1 also includes the frequency of reassessments (Freq) in the district. The period the data cover, 2001 to 2006, was a period of generally increasing property values. By law, property is required to be assessed at true market value if it were to be sold at the time of assessment. Additionally, the districts must comply with the state law that requires a minimum sales ratio of 0.70. However, since the majority of districts do not reassess on an annual basis, the assessors may take into account the time lag between reassessments with forward looking expectations to help remain in compliance with state law. That is, if they do not plan to reassess for another few years, during which property values appreciate, then they may overassess the property in anticipation.

Finally, a dummy variable indicating whether or not the assessor of the district is elected or appointed is also included in the regression. While voters can, and frequently do, voice their complaints about assessors to the elected officials who appointed them, it may still buy the assessment process more insulation from political pressure as compared with an elected assessor. Also, an indicator variable will be used to control for whether or not the year of the assessment coincided with an election year for that district.

After estimating Equation 4.1, the model will be extended to include a series of interaction terms between the other variables and the elected assessor dummy. This will provide insight as to whether differences in assessor types are being transmitted through their different responses to characteristics.

5. Estimation and Results

Fixed Effects Variance Decomposition

Researchers have long employed the fixed effects model in panel data because it carries several distinct advantages, arguably the most important of which is the ability to control for unobserved heterogeneity across areas. The major drawback of this procedure is that, since the correlation coefficients rely on within-variance, the time-invariant variables are not estimable. This is problematic for researchers interested in time-invariant variables who wished to also control for unmeasurable heterogeneity across areas. The Hausman-Taylor model is the traditional approach, but it is very demanding in terms of required instruments for proper estimation. An alternative approach developed by Plumper and Troeger (2007) is employed in this article, and since it is relatively new, a brief summary follows.

The true data generating process in a fixed effects model with time-invariant variables is


Equation 5.1 includes time-varying ([]) and time-invariant ([z.sub.i]) variables in addition to a Gaussian error term ([[epsilon]]) and unit-specific effects ([u.sub.i]). The fixed effects regression model is achieved by subtracting the time average of the variables from Equation 5.1, represented as


where [[??]] = [] - [[bar.y].sub.i], [[??]] = [] - [[bar.x].sub.i], and [[??]] = [] - [[bar.e].sub.i]. This regression in Equation 5.2 is estimated with OLS to obtain estimates of [beta], which will be labeled [[beta].sub.LS]. With these estimates, unit effects are calculated as

[[??].sub.i] = [[bar.y].sub.i] - [K.summation over (k = 1)] [[beta].sup.LS.sub.k][[bar.x]] - [[bar.e].sub.i]. (5.3)

As a result, [[??].sub.i] contains the unobserved unit-specific effects and the observed unit-specific effects. Next, [[??].sub.i] is regressed on the observed time-invariant variables in z, which breaks the unit effects into variance explained by z and unexplained variance, [h.sub.i]

[[??].sub.i] = [M.summation over (m = 1)] [[gamma].sub.m][z.sub.mi] + [h.sub.i]. (5.4)

This unexplained variance in [h.sub.i] resulting from the OLS of Equation 5.4 is then used as an estimator in the following regression that can be estimated by pooled OLS:


Plumper and Troeger (2007) correct the standard errors using a demeaned variance-covariance matrix and adjust the degrees of freedom to include the number of unit-specific effects. They further demonstrate how the model can be corrected for heteroskedasticity and a MA1 process, and they provide a series of Monte Carlo experiments that demonstrate the variance-decomposition model to have more reliable estimates than the Hausman-Taylor model, pooled OLS, and the random effects model.

Estimation Results

As previously described, the regression model attempts to explain variation in the median sales ratio with variables that might influence an assessor's behavior as they try to maximize political support. Table 3 provides the results of the model in Equation 4.1. OLS was used to estimate regressions without district fixed effects, but the fixed effect variance-decomposition (FEVD) method developed by Plumper and Troeger (2007) and described in the previous subsection was employed when district fixed effects were included. Variable definitions and sources can be found in Table 1. Hausman tests supported the use of an area fixed effects approach over a random effects model, and F-tests supported the inclusion of the district fixed effects over the OLS models without.

While the aforementioned statistical tests and goodness-of-fit measures support the models with district and time fixed effects, Table 3 nevertheless presents all the estimated models. While the descriptive statistics demonstrate that property is generally underassessed in all areas, it appears that having an elected assessor is correlated with a statistically significant lower level of assessment. The FEVD regressions revealed that having an elected assessor resulted in a -0.16 to -0.23 reduction in the sales price ratio, with and without time fixed effects, respectively. The signs of the other variables follow the expectations described in the previous section. To gain a greater understanding of how assessor type may be influential, Table 4 extends the FEVD model with time and district fixed effects to include interaction terms between the elected assessor dummy and the other independent variables. (14)

The first column of Table 4 demonstrates that once the interaction effects are included, having an elected assessor has a much smaller effect and is not statistically significant. Several of the interaction terms are statistically significant, however, which suggests that the significance of how elected assessors behave is actually channeled through a differential treatment of other factors. The last two columns of Table 4 demonstrate the marginal effect of the independent variables conditional on whether the assessor is elected or appointed. These marginal effects are for a 1 SD increase in the corresponding continuous independent variable and a discrete change for dummy variables. Interpreting the results in Table 4, it is seen that a 1 SD increase in the proportion of commercial property is correlated with a 0.022 increase in a district's sales ratio when the assessor is elected, but only a 0.003 increase when the assessor is appointed. The elected assessor's increase is statistically significant at the 5% level, and from the descriptive statistics in Table 2 it can be seen that this change represents about a quarter of a SD increase in the sales ratio (0.086). So the results would suggest that elected assessors tend to assess commercial property at higher rates relative to their residential counterparts, consistent with a practice of tax exporting.

Both elected and appointed assessors respond to increases in fiscal stress by raising assessments by a margin that is statistically significant at the 1% level. A 1 SD increase in the fiscal stress composite index is correlated with an increase of the sales ratio by 0.084 for appointed and 0.080 for elected assessors, which is nearly a SD of the dependent variable in magnitude. For the unemployment rate, the effect is statistically insignificant when the assessor is elected and only significant at the 10% level for appointed assessors. Analytically, the marginal effect of increasing the unemployment rate by 1 SD (1.491) is correlated with a 0.011 increase in districts with an appointed assessor but essentially has no effect for elected assessors. As previously discussed, greater unemployment is indicative of greater fiscal stress in the area, but the unemployed could be more likely to pressure officials for lower assessments. One way then to interpret the marginal effects of unemployment in Table 4 is that this latter effect is stronger among elected assessors than their appointed counterparts. ASD increase in median household income ($9530) is correlated with a -0.058 and -0.018 decrease in the sales ratio for appointed and elected assessors, respectively. For appointed assessors, the effect is statistically significant at the 1% level. A similar account occurs for racial homogeneity, in which a 1 SD increase in the index would result in a positive marginal effect for assessors of both types but is statistically significant only for appointed assessors. Racial homogeneity appears analytically significant in both assessor types, with even the marginal effect for the elected assessor being correlated with an increase of more than 1 SD in the sales ratio.

For both elected and appointed assessors, the marginal effect of being in an election year is statistically insignificant, though positive for appointed (0.016) and negative for elected (-0.022). The proportion of the population over the age of 65 is negative and statistically significant at the 1% level for only the appointed assessors while seemingly having no effect on elected assessors. The results also indicate that each additional year of lag between reassessments is correlated with a 0.01 increase in the sales ratio, implying that assessors anticipate housing price appreciation with higher assessments.

While the dummy variable for having an elected assessor carried a statistically insignificant coefficient of -0.020 in Table 4, an estimate of the effect of switching from an appointed assessor to an elected assessor would need to account for the cumulative effect of the variables interacting with the elected dummy. If the continuous independent variables take their corresponding mean sample value, the cumulative change in the predicted value of switching from an appointed to an elected assessor would correspond to a -0.135 decline in the median sales ratio of the district in a nonelection year and a -0.157 decrease during election years. Since this represents a nearly 2 SD decline in the sales ratio, the results are analytically significant. Since assessments are already low across assessor types, this finding supports the case that whatever political insulation from voters that appointed assessors enjoy results in more accurate assessments, ceteris paribus. This is also supportive of the Alesina and Tabellini (2008) theoretical model that concludes that public officials in positions requiring more technical skills should be appointed rather than elected.

Since the histogram in Figure 2 reveals the presence of three outliers considerably lower than the rest of the sample, a robustness check is in order. To determine how influential they were in the regression results, they were excluded from the sample and the coefficients estimated again according to the model presented in Table 4. Qualitatively, the results change very little. The largest effect was to increase the magnitudes of the coefficients for the unemployment rate as well as its corresponding interaction term with the dummy variable. Specifically, the unemployment rate coefficient increased from 0.007 to 0.017, and its interaction effect decreased from -0.008 to -0.015. The change in predicted value from switching between appointed to elected also became slightly more negative, which is not surprising since the three lowest observations were counties with appointed assessors. (15)

6. Discussion

While the task of assessing property is a job of measurement, the assessor's office is likely subject to political pressures from property owners because they are either elected themselves or appointed by another elected official. In addition, they likely desire support from other elected officials for their own reelection, for reappointment, or for budgetary reasons. This article has empirically tested the influence of various sources of this pressure and their implications for the assessment-to-sale price ratio (sales ratio) using data from Virginia cities and counties over the 2001 to 2006 period. As state governments become more involved in the administration and collection of property taxes from local officials, the accuracy of the sales ratio becomes increasingly important.

The findings are generally consistent with the view from the ACIR (1963) that both appointed and elected assessors would face similar pressures. The results show that the direction of the effect for most of the political pressure variables is the same, though differences often exist in terms of magnitude and statistical significance. For instance, both assessor types assess commercial property at higher rates, but the marginal effect is much larger and only statistically significant for elected assessors. Since commercial property is more likely to be owned by nonresidents, this behavior is consistent with tax exporting. Furthermore, greater levels of fiscal stress were positively correlated with higher sales ratios, while the property tax rate was negatively correlated.

Socioeconomic variables of the population, such as racial homogeneity, income, and share of the population over age 65, were also statistically significant for appointed assessors. Elected assessors were less responsive in terms of marginal effects on these dimensions, and their coefficients were statistically insignificant. Racial homogeneity was the most analytically influential of these, since a 1 SD increase in homogeneity was correlated with a more than 1 SD increase in the sales ratio for both assessor types. One explanation for this presented in the article is that it is easier for assessors to identify the median voter in racially homogeneous districts, thereby allowing them to overassess other groups. Finally, the results also indicate that assessors are forward looking when property is not reassessed on an annual basis, so they account for housing price appreciation with higher levels of assessment.

Cumulatively, the effect on the predicted value of switching from an appointed assessor to an elected one is estimated to decrease the sales ratio by -0.135 in nonelection years and -0.157 in election years. Considering that the sample SD of the sales ratio is 0.086, this result is analytically significant. Since property is already underassessed by both types, this estimate infers that appointed assessors are actually more accurate. These results are consistent with the conjecture of Alesina and Tabellini (2008) that public officials in positions requiring more technical skills should be appointed rather than elected.

By identifying these sources of local political pressure, states may be better able to monitor the assessors to prevent biased assessments. In particular, the degree of separation between the public and the appointed assessor must allow for enough political autonomy to result in more accurate assessments. The results here suggest that reforms aimed at moving assessors off the voter ballot could improve assessor accuracy.

Virginia is unique in that they have a mixed system of elected and appointed assessors, but this uniqueness may limit the application of the results to other states. Future research could attempt to control for the various institutional differences that exist among the states in their administration of property tax assessments and determine the extent of political influence on the assessors themselves. (16) Furthermore, the time period of the data was one of rising property values in Virginia. In future data releases it could be possible to study the effects in an environment of declining housing prices.


Advisory Commission on Intergovernmental Relations (ACIR). 1963. The role of the states in strengthening the property tax. Volume 1. Washington, DC: U.S. Government Printing Office.

Alesina, A., R. Baqir, and W. Easterly. 1999. Public goods and ethnic division. Quarterly Journal of Economics 114:1243-84.

Alesina, A., and G. Tabellini. 2008. Bureaucrats or politicians? Part II: Multiple policy tests. Journal of Public Economics 92:426-47.

Belsley, D. A., E. Kuh, and R. E. Welsch. 1980. Regression diagnostics: Identifying influential data and sources of collinearity summary. New York: Wiley.

Besley, T., and S. Coate. 2003. Elected versus appointed regulators: Evidence and theory. Journal of European Economic Association 1:1176-206.

Black, D. E. 1977. Property tax incidence: The excise-tax effect and assessment practices. National Tax Journal 30:429-34.

Bowman, J. H. 2006. Property tax policy responses to rapidly rising home values: District of Columbia, Maryland, and Virginia. National Tax Journal 59:717-33. Bowman, J. H., and J. L. Mikesell. 1989. Elected versus appointed assessors and the achievement of assessment uniformity. National Tax Journal 42:181-9.

Cornia, G. C., and L. C. Waiters. 2006. Full disclosure: Controlling property tax increases during periods of increasing housing values. National Tax Journal 59:735-49.

Fischel, W. A. 2001 a. The homevoter hypothesis: How home values influence local government taxation, school finance, and land-use policies. Cambridge, MA: Harvard University Press.

Fischel, W. A. 200lb. Homevoters, municipal corporate governance, and the benefit view of the property tax. National Tax Journal 54:157-74.

Fisher, G. W. 1996. The worst tax? A history of the property tax in America. Lawrence, KS: University Press of Kansas.

Fiva, J. H., and M. Ronning. 2008. The incentive effects of property taxation: Evidence from Norwegian school districts. Regional Science and Urban Economics 38:49-62.

Gardner, A. 2006. Northern Virginians greet assessments with disbelief. Washington Post, 6 March, B01.

Geist, W. E. 1981. Realty taxes ignite protest in Jersey. The New York Times. 1 September. Accessed 22 June, 2007. Available http://query.nytimes.comL

Glaeser, E. L. 1995. The incentive effects of property taxes on local governments. NBER Working Paper 4987.

Hanssen, A. 1999. The effect of judicial institutions on uncertainty and the rate of litigation: The election versus appointment of state judges. Journal of Legal Studies 28:205-32.

Hoover, G. A. 2008. Elected versus appointed school district officials: Is there a difference in student outcomes? Public Finance Review 36:635-47. International Association of Assessing Officers (IAAO). 1999. Standard on ratio studies. Assessment Journal 6:23-55.

Johnson, M. S. 1989. Assessor behavior in the presence of regulatory constraints. Southern Economic Journal 55:880-95.

Joravsky, B. 2005. Is a wink as good as a nod? Chicago Reader. 8 September. Accessed 22 June, 2007. Available

Lind, J. T. 2007. Fractionalization and the size of government. Journal of Public Economics 91:51-76.

Lowery, D. 1982. Public choice when services are costs: The divergent case of assessment administration. American Journal of Political Science 26:57-76.

Lowery, D. 1984. Tax equity under conditions of fiscal stress: The case of the property tax. The Journal of Federalism 14:5545.

Payton, S. 2006. A spatial analytic approach to examining property tax equity after assessment reform in Indiana. The Journal of Regional Analysis and Policy 36:182-93.

Plumper, T., and V+ E. Troeger. 2007. Efficient estimation of time-invariant and rarely changing variables in finite sample panel analyses with unit fixed effects. Political Analysis 15:98 114.

Stewart, D. O. 1977. The census of governments' coefficient of dispersion. National Tax Journal 30:85-8.

Strauss, R. P., and S. R. Sullivan. 1998. The political economy of the property tax: Assessor authority and assessment uniformity. Presented at National Tax Association 91st Conference.

Virginia Department of Housing and Community Development (VDHCD). 2006. Report on the comparative revenue capacity, revenue effort, and fiscal stress of Virginia's counties and cities 2003/2004. Richmond, VA: VDHCD.

Virginia Department of Taxation (VDT). 2007. The 2005 Virginia assessment/sales ratio study. Richmond, VA: VDT.

(1) See The Worst Tax? A History of the Property Tax in America by Fisher (1996) for arguments as to why property taxes carry so much disdain.

(2) For examples in the news, see Gardner (2006), Joravsky (2005), and Geist (1981) for coverage of home owners pressuring the assessor's office in periods of rising property values.

(3) The IAAO standards require that assessors error on average by no more than 10% above or below fair market value in their assessments (IAAO 1999).

(4) Virginia, the state studied in this article, underwent such a reform in 1975 and therefore predates the dataset in this article by 26 years.

(5) The 12 states not currently collecting property tax revenue at the state level include Colorado, Connecticut, Delaware, Hawaii, Idaho, Iowa, New York, Oklahoma, South Dakota, Indiana, Texas, and Utah.

(6) This insight of Glaeser (1995) has drawn empirical support from Fiva and Ronning (2008) because of its implications for encouraging competition among schools.

(7) It appears from observation of the tables presented in Strauss and Sullivan (1998) that states with predominantly elected assessors have more accountability standards, but I will leave that issue to future research.

(8) I use the term "district" to refer to Virginia's counties and independent cities.

(9) Section 58.1-3259 of Virginia Tax Code.

(10) For recent examples of fractionalization in the public economics literature, see Lind (2007) for an analysis on higher fractionalization leading to lower redistribution levels and Alesina, Baqir, and Easterly (1999) for its influence in reducing the provision of public goods.

(11) The three excluded districts are Albemarle County, Henrico County, and Prince William County. Albemarle's Board of Supervisors appoints the County Executive Officer, who in turn appoints the assessor. Henrico and Prince William's Board of Supervisors directs the Department of Finance, which includes the Real Estate Assessment Office.

(12) Those results are available upon request.

(13) A complication arises with using the fiscal stress index, since it is at least partially collinear with the district's income and property taxes. To separate these effects, the reported fiscal stress score is regressed on income and property tax rate. The residuals of this regression then represent the fiscal stress in a district unexplained by and orthogonal to income and property taxes. Thus the fiscal stress variable ultimately used in the estimation of the final results will be the residuals from this regression. The regression results for the fiscal stress regression on income and property tax rate are not reported, but they are available upon request.

(14) The interaction term with the frequency of reassessment variable (Freq) was omitted because its p value was 0.99 and excluding it made no noticeable effect on the remaining estimates.

(15) These results are not printed here, but they can be supplied upon request.

(16) For example, Maryland's local assessors are appointed by the state, which might undermine local political pressures.
Table 1. Variable Definitions and Source

Variable Name                               Definition

1. Sales ratio (a)          The median sales ratio of a
                              jurisdiction: in each district, a sample
                              of fair market transactions were taken
                              and divided into the assessed value on
                              the books at the time of the sale
2. Elected (b)              Indicator variable where a jurisdiction
                              with an appointed assessor takes a zero
                              value, else takes a value of one
3. Election year            Indicates that the year of observation
                              is an election year for that
4. Racial homogeneity (c)   Author's calculation of =
                              [[(%White).sup.2] + [(%Black).sup.2] +
                              [(%Asian).sup.2] + [(%Indian).sup.2] +
                              [(%Miscellaneous).sup.2]] (1/10,000)
5. Share age 65+ (c)        Residents over the age of 65 as a
                              proportion of the total population
6. Income (d)               The median adjusted gross income on all
                              state tax returns in thousands of
                              dollars for year of regression
7. Pop growth (c)           Average annual population growth from
                              previous year
8. Share commercial (a)     The number of property sales classified
                              as commercial divided total number of
                              sales of all property types
9. Property tax rate (a)    Property tax rate levied by the
                              district, typically determined by Board
                              of Supervisors
10. Fiscal Stress (d)       The residuals from the regression of
                              fiscal stress index on income and
                              property tax rate: the fiscal stress
                              index is intended to measure the
                              budgetary pressure facing the district's
                              government; regression results available
                              upon request
11. Unemployment Rate (e)   The percentage unemployment rate in the
12. Freg (b)                The number of years between
                              reassessments authorized by the
                              jurisdiction; for example, a
                              jurisdiction with annual reassessments
                              would have a value of zero

(a) VDT (2007).

(b) Author's research with aid of Virginia Association of
Assessing Officers jurisdiction directory (

(c) U.S. Census Bureau.

(d) VDHCD (2006).

(e) U.S. Bureau of Labor Statistics.

Table 2. Descriptive Statistics

                     Mean    Deviation     Min      Max

Sales ratio          0.843     0.086       0.380    1.010
Elected              0.407     0.492       0.000    1.000
Election year        0.253     0.435       0.000    1.000
  homogeneity        0.662     0.151       0.436    0.986
Percent age 65+      0.131     0.042       0.043    0.285
Income              30.693     9.530      17.420   66.211
Annual pop
  growth             0.010     0.018      -0.029    0.093
  commercial         0.031     0.025       0.000    0.150
Property tax rate    0.894     0.291       0.290    1.450
Fiscal stress        0.188     6.610     -36.929   20.594
  rate               3.974     1.491       1.600   12.500
Freq                 2.364     1.707       1.000    6.000

Table 3. Unbalanced Panel Regression Results
for 2001-2006 Virginia Cities and Counties

Dependent Sales Ratio                 OLS

Elected                  0.028 (**)        0.038 (***)
                        (0.014)           (0.014)
Election year            0.004             0.006
                        (0.010)           (0.010)
Racial homogeneity       0.133 (***)       0.092 (***)
                        (0.037)           (0.035)
Share age 65+           -0.450 (***)      -0.062
                        (0.157)           (0.156)
Income                   0.000             0.003 (***)
                        (0.001)           (0.001)
Annual pop growth       -0.632 (**)       -1.022 (***)
                        (0.310)           (0.265)
Share commercial         0.537 (***)       0.475 (**)
                        (0.203)           (0.203)
Property tax rate        0.048 (*)         0.028
                        (0.027)           (0.026)
Fiscal stress            0.000             0.003 (***)
                        (0.001)           (0.001)
Unemployment rate        0.005             0.007 (*)
                        (0.004)           (0.004)
Freq                     0.005             0.000
                        (0.006)           (0.005)
Intercept                0.687 (***)       0.561 (***)
                        (0.082)           (0.084)
District fixed effect   No                No
Year fixed effect       No               Yes
[R.sup.2]                0.196             0.362

Dependent Sales Ratio                 FEVD

Elected                  -0.231 (***)      -0.160 (***)
                         (0.019)           (0.014)
Election year             0.002             0.006
                         (0.009)           (0.009)
Racial homogeneity        1.583 (***)       1.033 (***)
                         (0.089)           (0.055)
Share age 65+            -1.492 (***)      -0.373 (***)
                         (0.153)           (0.124)
Income                   -0.016 (***)      -0.004 (***)
                         (0.001)           (0.001)
Annual pop growth        -0.118            -0.126
                         (0.250)           (0.217)
Share commercial          0.325 (*)         0.393 (***)
                         (0.189)           (0.145)
Property tax rate        -0.342 (***)      -0.383 (***)
                         (0.031)           (0.027)
Fiscal stress             0.004 (***)       0.012 (***)
                         (0.001)           (0.001)
Unemployment rate        -0.010 (***)      -0.001
                         (0.004)           (0.003)
Freq                      0.040 (***)       0.018 (***)
                         (0.005)           (0.004)
Intercept                 0.604 (***)       0.434 (***)
                         (0.049)           (0.043)
District fixed effect   Yes               Yes
Year fixed effect        No               Yes
[R.sup.2]                 0.717             0.808

Sample size is 324. Robust standard errors (clustered at district
level) reported in parentheses. Statistical significance from
zero indicated at 0.01 (***), 0.05 (**), and 0.10 (*) level.
Metropolitan statistical area fixed effects included in all
regressions with results available upon request.

Table 4. Unbalanced Panel Regression Results and Marginal
Effects for 2001-2006 Virginia Cities and Counties

       Dependent Sales Ratio
       (r = 0.843, s = 0.086)                    Marginal Effects

       Variable                 FEVD           Appointed     Elected

Elected                      -0.020 (0.083)                 -0.020
Election year                 0.016 (0.010)   0.016        -0.022
Racial homogeneity      1.271 (***) (0.079)   0.192 (***)   0.096
Share age 65+          -0.828 (***) (0.177)  -0.034 (***)   0.000
Income                 -0.006 (***) (0.001)  -0.058 (***)  -0.018
Annual pop growth             0.018 (0.258)   0.000        -0.013
Share commercial              0.113 (0.199)   0.003         0.022 (**)
Property tax rate      -0.409 (***) (0.031)  -0.119 (***)  -0.087 (***)
Fiscal stress           0.013 (***) (0.001)   0.084 (***)   0.080 (***)
Unemployment rate         0.007 (*) (0.004)   0.011 (*)    -0.001
Freq                    0.010 (***) (0.004)   0.018 (***)   0.018 (***)
Elected x election      -0.038 (**) (0.016)
Elected x racial       -0.631 (***) (0.066)
Elected x share age     0.838 (***) (0.241)
Elected x income        0.004 (***) (0.001)
Elected x pop growth     -0.753 (*) (0.443)
Elected x share comm     0.761 (**) (0.295)
Elected x prop tax      0.109 (***) (0.042)
Elected x fiscal             -0.001 (0.001)
Elected x unemp rate         -0.008 (0.007)
Intercept               0.430 (***) (0.052)

Sample size is 324 and [R.sup.2] = 0.82. Marginal effects are
based on a standard deviation change for continuous independent
variable, and discrete change for dummy variables. Robust
standard errors (clustered at district level) reported in
parentheses. Statistical significance from zero for correlation
coefficients indicated at 0.01 (***), 0.05 (**), and 0.10 (*)
level. Year fixed effects not reported but available upon

Justin M. Ross, Assistant Professor of Economics, School of Public and Environmental Affairs, 1315 E 10th Street, Indiana University, Bloomington, IN 47405, USA; E-mail

I appreciate the support of the West Virginia University Foundation through the Distinguished Doctoral Student Fellowship. I am extremely grateful to Santiago Pinto, George Hammond, Stratford Douglas, John Blair, Russell Sobel, Donald Lacombe, Cynthia Rogers, William Shughart, and Brian Cushing for their helpful comments and suggestions. Any mistakes are my own.

Received May 2009; accepted January 2010.
COPYRIGHT 2011 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Ross, Justin M.
Publication:Southern Economic Journal
Geographic Code:1USA
Date:Jan 1, 2011
Previous Article:What do OECD countries cut first when faced with fiscal adjustments?
Next Article:Factor returns and circular causality.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters