Aid and public finance: a missing link?
Findings in recent publications generally agree that foreign aid does not have the desired positive effect on growth and poverty alleviation. The two main reasons for this seem to be political patterns of aid-flows and poor performance of recipient countries' public institutions (Alesina and Dollar 2000).
We investigate the relationship between the Quality of Public Financial Institutions (PFI quality) of less developed countries, and the level of multilateral aid they receive. This raises two central sets of questions:
First, does PFI quality develop in reaction to multilateral aid received? If such a relationship exists, will it be positive or negative; i.e., does multilateral aid foster or hinder public finance quality and thereby good governance and institutional development?
Second, do multilateral aid-flows move relative to the recipient country's PFI quality? If so, does empirical evidence suggest that multilateral donors regard improvements in PFI quality as an incentive (or condition) to increase aid-giving? Or could there be the case that recipient countries with poor institutions attract more aid?
This paper uses empirical evidence, in the form of panel data on 86 countries over a 19 year period, to answer the questions above. Apart from a set of robust econometric specifications, our research also requires the construction of a coherent measure of PFI quality. For this purpose we set up a new Public Financial Institution Quality (PFIQ) Index, which draws from standardised assessments of institutional quality, as well as from reliably available national data.
We restrict our investigation to multilateral aid-giving, as its underlying decisions should be free from political and strategic motives (Alesina and Dollar 2000). Multilateral aid is defined as foreign aid donations from the following: African Development Bank, African Development Fund, Asian Development Bank, Asian Development Fund, European Commission, International Bank for Reconstruction and Development, International Development Association, InterAmerican Development Bank, Inter-American Development Bank Special Fund, International Fund for Agricultural Development, United Nations Development Program, United Nation's Children's Fund, Joint United Nations Programme on HIV/AIDS, United Nations Population Fund, and the Global Fund.
The bulk of existing literature on foreign aid is divided into two sections. One section studies the effects of foreign aid on the recipient country's institutions and governance, and the other studies the impact of changes in institutions and governance on aid-flow patterns.
Concerning the effects of foreign aid on recipient country's institutions, Boone (1996) argues that aid does not cause an increase in investment or improve human development indicators, but it does increase the size of government. Similarly, Burnside and Dollar (2000) estimate an equation for government consumption as a share of GDP, and find that aid has a strong positive impact on government consumption. Both studies suggest that aid does have an effect on governmental institutions, but that these changes, in turn, seem not to positively translate into growth and poverty alleviation. This puzzle suggests that the black box of governance in developing countries is not working effectively. This may be due to misguided incentives and rent-seeking behavior or to chronic institutional inabilities--both of which are related to PFI quality.
In their 2000 paper, Alesina and Dollar find that exogenous changes in aid have no impact on recipient countries' levels of democratization. On the same note, Burnside and Dollar (2000) argue that aid has no substantial effect on economic policies. Assuming that institutional variables such as democratization and economic policies are closely intertwined with PFI quality (because of positive externalities or spill-over effects), these findings suggest that aid may not have any significant impact on PFI quality.
Knack (2001) finds that higher aid levels erode the quality of governance, which he measures by bureaucratic quality, rule of law, and corruption. Alesina and Weder (2002) furthermore provide tentative evidence of increases in aid being correlated with increases in corruption. On a similar note, Brautigam and Knack (2004) find that higher levels of aid are associated with lower tax effort (defined as the ratio of tax revenue to GDP). These findings suggest a negative relationship between aid and PFI quality, of which governance quality and tax effort are indicative measurements.
Few findings do suggest a positive correlation between targeted foreign aid and government institution quality. In our case of interest, aid is sometimes used for improved training and increased salaries for public employees, including judges and tax collectors. Van Rijckeghem and Weder (2001) argue that because of the increase in salaries, more competent bureaucrats can be employed and bribe solicitation can be reduced, thereby dissuading rent-seeking behavior. The result will be an improvement in tax effort and government creditworthiness, which would both provide support for the country's PFI quality.
With respect to the impact of institutional and governance changes on aid, Alesina and Dollar (2000), and Knack (2001) point out that, at the margin, institutional factors and democratization do attract aid, even though political considerations on the donors' behalf do play a significant role. These results hint at a possible positive response of multi-lateral aid to PFI quality improvements.
An interesting, though contested, finding by Burnside and Dollar (2000) asserts that aid raises growth in a good policy environment. However, Easterly et al. (2004) show that these results are not robust to the use of additional data.
Despite questionings concerning the strength of their empirical findings, Burnside and Dollar's underlying theory finds resonance in other publications. Boone (1996), for example, shows that aid targeted to liberal regimes may be more successful in alleviating social misery. In a similar fashion,
Brautigam and Knack (2004) reason that Sub-Saharan Africa's development problems ultimately reflect a crisis of governance, and argue that only aid provided on a more selective basis can trigger a virtuous cycle of development.
On the assumption that multilateral aid-flows are purpose-bound and designed to serve effective growth rather than political strategies, the findings cited above suggest that multilateral aid should be attracted towards democratization, liberal regimes, and sound economic policies. Granted that these three variables are reflected in good PFI quality, donors should either reward or condition aid on improvements in recipient countries' public finance systems.
A puzzling result from Alesina and Weder (2002) is that multilateral donors seem not to take account of a recipient nation's degree of corruption when allocating aid. In contrast to this, Burnside and Dollar (2000) find that multilateral aid is allocated in favor of good policy. These contradictory findings, then, do not yield clear-cut assumptions on the relationship between PFI quality and aid.
The next section argues why we chose to focus on public finance systems; thereafter, we discuss the empirical equations we put to use and our methods of investigation. The section that follows explains how we constructed the PFIQ Index. This section is then succeeded by a summary of our data sources and our methods of compilation. Based on this, we outline the various regression specifications and the identification strategies used, and then we follow with sections that detail the results for PFIQ regressions and aid regressions. Finally, we provide a summary of caveats and a conclusion.
The Case for Public Finances
The quality of public finance is determined by the level and composition of public expenditure and its financing via revenue and deficits. An ECB working paper (Afonso et al. 2005) finds that high quality public finances support the long-run growth of the economy. Specifically, low deficits and debt ratios create expectations that public finances are sustainable, so that expenditure policies and tax systems and rates will be predictable. This supports economic growth because it creates an environment conducive to long-term oriented savings and investment decisions. By contrast, if, over a longer period, government revenue is much lower than total public spending--creating unsustainable macro imbalances and public debt accumulation--growth may be reduced because the private sector might come to see the fiscal situation as unsustainable and reduce investment in anticipation of future higher taxes.
Public finance systems have received increased attention in the donor community. In 2005, more than 100 countries and donor organizations recognized the imperative of managing aid more rationally when they endorsed the Paris Declaration on Aid Effectiveness. A cornerstone of the plans to reform the system of aid delivery was to strengthen and use country systems, that is; use the national government circuits of public finance rather than by-pass these circuits to deliver aid to citizens. Donors agreed to concede that the degree to which they relied on recipient country's public finance systems depended at least partly on the quality of these systems, and that progress will depend on greater understanding of the development benefits and risks of using these systems, l The motivation for this paper is inscribed directly in the line of this last citation.
Despite this increased interest, no index in the academic circles has yet been constructed. In parallel to the Paris Declaration, there have been such efforts in the policy-making arena. The outcome has been the Public Expenditure and Financial Accountability (PEFA) framework (2005). Unfortunately, the data from this collaboration is very scarce and in any case only goes back to 2006 when the first assessment reports were drafted.
Another relevant measure of the quality of the public finance system would be the indicator 13, Quality of Budget and Financial Management, of the Country Policy and Institutional Assessment ratings, produced by the World Bank. The problem with this indicator is twofold. Data for the CPIA ratings have only been made public from 2004 onwards; and, the CPIA ratings form the basis for the World Bank's allocation of aid, which implicates issues of political controversy we seek to avoid.
Empirical Equations and Methods of Investigation
Our empirical work tries to answer the following two questions: Do public financial systems develop in reaction to multilateral foreign aid received? Do aid-flows move relatively to the quality of the recipient's public financial system? We investigate these questions by estimating variants of the following two equations:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where i indexes countries, t indexes time, [PFIQ.sub.(i, t)] is the Quality of Public Financial Institutions index score, [a.sub.(i, t)] is aid receipts, [z.sub.(i, t)] is a Z x 1 vector of other control variables that might affect the Public Finance Quality index score, [p.sub.(i, t)] is a P x 1 vector of other exogenous variables that might affect the allocation of aid, [y.sub.i, t]) is the logarithm of initial real per capita GDP, [[epsilon].sub.(PFQ)] (i, t) and [[epsilon].sub.(a) (i, t)] are error terms.
The main econometric difficulty related to work on aid and policies is that the error terms in Eqs. 1 and 2 are possibly correlated. In accord with many other studies (e.g. Burnside and Dollar 2000; Alesina and Weder 2002; Knack 2001; Boone 1996; Brautigam and Knack 2004), we intend to tackle this problem of reverse causation, using a two-stage-least-squares (2SLS) instrumental variable procedure.
The Public Finance Institutions Quality (PFIQ) Equation
On the basis of the literature and theory outlined in previous sections, we allow for the quality of a country's public finance system to depend on the level of aid received. We also assert that the level of public finance quality is dependent upon the level of income of the country, hence we include y(i, t), the logarithm of initial GDP per capita. The set of included covariates (vector Z) is comprised of:
Volume of financial system, proxied by the level of broad money (King and Levine 1993);
Political violence, which is measured by military in politics, internal conflict and external conflict measures of the International Country Risk Guide (ICRG); and Macroeconomic policies, which is composed of year-on-year inflation rate.
The Aid Equation
We rely on the literature and theory cited above to determine the variables which would affect the allocation of aid. Obviously, central to our paper is the quality of public finance systems of the recipient country, so this is included in the equation. Based on our literature review we consider multilateral aid to be outcome-bound, rather than driven by political considerations, which is why we do not control for political strategies and post-colonial alliances. Included covariates (vector P) are:
--The logarithm of initial GDP per capita [y.sub.(i, t)], and the logarithm of population;
--A control variable for economic policies, measured by an index of economic openness (computed as the ratio of exports and imports over GDP);
--The democratic nature of the recipient country, indicated by a democracy dummy variable (based on Polity IV index), will also be added to the model, since multilateral donors may discriminate against non-democratic regimes;
--Measures of recipient need: Infant mortality rate, and Illiteracy rate.
Constructing the Public Finance Institutions Quality (PFIQ) Index
The quality of a country's public finance system is determined by the level and composition of public expenditure and its financing via revenue and deficits (Afonso et al. 2005). There are five sub-components of public finance which have to be considered when assessing a country's public finance system: (i) the strength of the institutional framework in which the system is embedded; (ii) the level of governmental corruption; (iii) the robustness of the legal framework; (iv) the efficiency of the tax systems; and, (v) the contribution to a macroeconomic environment which is stable and conducive to private sector participation. We chose the following proxies for these sub-components:
(i) Institutional strength (IS): productive output of a public finance system is highly sensitive to the strength and expertise of the bureaucracy to govern without drastic changes in policy or interruptions in government services. If the bureaucracy can act as a shock absorber when governments change, then the budget will not be subject to frequent, unpredictable and erratic modifications. As such, the strength of the bureaucracy underpins the stability and predictability of the public finance system. The strength of the bureaucracy is measured by the Bureaucracy Quality rating, a component of the Political Risk measure constructed by the International Country Risk Guide (ICRG).
(ii) Enforcing legal/framework (LF): An enforcing legal framework will dissuade rent-seeking behavior. Enforcing is understood not only in the sense of an impartial legal system, but also in the sense of a popular observance of the law. The conjunction of a strong judiciary system and a belief in the enforcement of the law provides strong disincentives for rent-seeking behavior, and thereby strengthens the public finance system. The legal framework is assessed by the Law and Order rating, another component of the Political Risk measure constructed by the International Country Risk Guide (ICRG).
(iii) Governmental corruption (GC): Corruption is a particularly important feature of problems related to public finance quality, and as such, it has to be dissociated from the institutional and the legal framework. Corruption captures the close ties between government officials and businesses, and these ties can distort the distribution of government allocations. Moreover, corruption can shift government spending from productive policies to policies favoring smaller groups and may impede the efficiency of tax assessments. All of these factors contribute directly to a deteriorating public finance system. The greater risk is that corruption at the governmental level can become so overwhelming it results in a fall or major restructuring of the governmental institutions, thereby undermining the long-term credibility of the public finance system. The level of governmental corruption is assessed by the Corruption rating, a further component of the Political Risk measure constructed by the International Country Risk Guide (ICRG).
(iv) Tax collection (TC): An important feature of a public finance system, TC is the capacity and the efficiency by which the government collects taxes. A strong tax collection system not only increases tax revenue, thereby minimizing the risk of running a government deficit, but it also, specifically in the case of developing countries, reduces the country's reliance upon other sources of funding (such as foreign aid). Tax effort is defined as the ratio of tax revenue to GDP (Brautigam and Knack 2004). The transfers received through tax revenue exclude grants from foreign governments and from international organizations.
(v) Contributfon to macroeconomic stability and conducive private sector participation (SP): As argued earlier (III), the quality of a public finance system depends in large part on its sustainability and predictability. Such two factors are best accounted for by the government budget deficit and surplus. Sporadic and large changes oscillating between surplus and deficit reflect the government's inability to manage resources on a longer-term basis and provide an insecure context for donors to inject money and for private agents to participate. A persistent high deficit reflects the government's unsustainable level of spending, undermining the authority and credibility of the country's public finance systems. Government surplus and deficit are measured as the difference between total tax revenue (excluding foreign aid and other grants) and total government expenditure, divided by GDP.
We give equal weight to each of the five subcomponents as we have not yet encountered any theoretical reason to do otherwise. The Quality of Public Finance Institutions (PFIQ) Index is therefore computed as follows:
[PFIQ.sub.(i, t)] = [[gamma].sub.1]*[IS.sub.(i, t)] + [[gamma].sub.2]*[LF.sub.(i, t)] + [[gamma].sub.3]*[GC.sub.(i, t)] + [[gamma].sub.4]*[TC.sub.(i, t)] + [[gamma].sub.5]*[SP.sub.(i, t)],
where i indexes countries, t indexes time; [summation] [[gamma].sub.i] = 1,...,5; IS = (Bureaucracy Quality Rating/4) *100; LF = (Law and Order rating/6) * 100; GC = (Corruption Rating/6)* 100; TC = (Tax Revenue, constant US$2000/GDP, constant US$2000)*100; SP = ((Total Tax Revenue, constant US$2000--Total Government Expenditure, constant US$2000)/ GDP, constant US$2000)*100.
The overall PFIQ score will be a percentage, which has no absolute, but only a relative value. The interpretation of the index is straightforward--a higher percentage score indicates a higher level of quality of the public finance system.
Data Sources and Compilation
In compiling our dataset, we collected data for each of the below variables (See Table 1) for all available countries and years. In most cases, we retrieved the data online, downloading it from the respective provider's webpage; only the ICRG measure we received from our LSE's Data Library.
Our values of interest were generally easily available, excluding M2 (brad money) and Tax Revenue, where we had to make the following adjustments. For M2, we downloaded the Country Table from IMF International Financial Statistics for all countries, extracted and summed the lines 14A.ZF, 24..ZF and 25...ZF, scaled all values to billions and thus compiled a single table measuring M2 for 185 countries from 1948 to 2009. In the case of Tax Revenue, the values for the years from 1990 onwards, were easily available in a single table for all countries. The historical values (1972-1989) we had to collect country by country, as there was no cross-country table available. Once the complete table was set up, we converted the values for each country into constant 2000 USD, using the Penn Exchange Rate 2000.
After we collected the complete dataset as outlined above, we reduced it to the panel that was covered by all variables, leaving us with 86 countries and a timeframe from 1984 to 2003. In the course of our empirical analysis, we also compiled two balanced datasets. First, we noticed that our PFIQ Index measure was considerably disturbed by a small number of outliers. While the vast majority of PFIQ Index values was smoothly distributed between (-1) and (+4), there were 14 observations ranging from (+18) to more than (+600). These measures can be attributed to measurement error in the collected datasets. We therefore decided to create a clean measure of PFIQ, by limiting the PFIQ values used in STATA to those below (+4). The resulting variable is called PFIQClean. Secondly, we compiled a reduced dataset, in which we only included those countries for which the PFIQ-relevant data were mostly complete; we did this with the intention to check whether our results from the big dataset were skewed by gaps in the data.
Summary of Regression Specification and Identification
We estimate variants of both equations. We have no prior theoretical basis which would allow us to discriminate between the use of either
aid/GDP or aid/population. Running all regressions with both variables, we find that for the majority of our results, the amplitude of the coefficients change, but the significance and the direction of the impact are consistently the same using both aid/GDP and aid/ population.
Concerning the PFIQ score, we estimate coefficients with and without the outliers from the PFIQ variable. The observations identified as outliers are caused by missing data in computing the PFIQ score, not by genuinely abnormally high PFIQ scores. We refer to Table 2 for a summary of the variables included in our regressions.
We mainly rely on the theoretical assumptions outlined in previous subsections to argue for the validity of our instrumental variables. To instrument for measures of aid in the PFIQ equation, illiteracy, infant mortality, and logarithm of population turned out to best satisfy the requirements for a good instrument. For the 2SLS aid regressions we decided to use internal conflict, external conflict and military in politics to instrument for PFIQ measures.
Public Finance Institutions Quality Regressions
Overall, our findings provide strong support for the hypothesis that aid fails to have a positive impact on PFIQ. In the unbalanced sample, both OLS and IV regressions yield an insignificant impact of aid on PFIQ. The sample without outliers and the PFIQ balanced sample both yield a significant, negative coefficient for the aid- variable. Differences in the coefficients between aidP and aidPC were negligible. In both balanced samples, OLS and IV coefficients are significantly different (as suggested by the Durbin-Wu-Hausman test), and we rely on the IV coefficients to support our results.
For the unbalanced sample, the first two regressions (Table 3, Eqs. 1 and 2) produce insignificant coefficients for almost all variables. We seek out to find relevant instruments for the aid-variable, in order to assess whether the insignificance is due to bias or indeed a non-existing relationship. We argued earlier that possible candidates for instruments should first be found in the control variables in the aid equation. Computing the correlation of aid per capita and aid over GDP with the candidate variables, we decided to use illiteracy and infant mortality rate as instruments for aid over GDP, and log of population as instrument for aid per capita (the correlation values are available on request).
Running the two 2SLS regressions (Table 3, Eqs. 3 and 4), we find that IV coefficients do not seem to differ from the OLS results determined above. This increases confidence in our OLS estimates, which show an insignificant coefficient for the aid-variable in the (unbalanced) sample with all observations.
We do, however, find significant results once we exclude outliers (there are 14 outliers--see section on data sources--out of 870 observations). Regressions without outliers both in the unbalanced and the PFIQ balanced sample produce the same result; multilateral aid has a significant and negative impact on our measure of public finance quality.
Proceeding as with the unbalanced sample, we first run the general regressions (Table 3, Eqs. 5 and 6). The aid coefficient is significant at the 10% level and the 5% level, respectively. Furthermore, the overall explanatory power of our specification has greatly increased. R-squared has gone from 0% in the unbalanced regressions to roughly 58% in both Eqs. 5 and 6. Finally, all control variables, except the logarithm of GDP per capita, have become highly significant.
We then investigate the relationship using 2SLS. Using illiteracy and mortality as instruments for aid over GDP (despite the somewhat significant relationship between mortality and PFIQClean), we find the coefficient of aid over GDP is negative and significant at the 1% level.
For aid per capita, using logarithm of population as an instrument (Table 3, Eq. 7), we also find a highly significant negative coefficient. Adding infant mortality rate as an instrument only slightly increases the significance of the aid coefficient.
To test whether the IV and the OLS coefficients are significantly different, we resort to the Durbin-Wu-Hausman test (Table 3, Eqs. 7 and 8). Yielding a chi- square test-statistic of 12.05 and 41.4 for aid per capita and aid over GDP, respectively, we conclude that OLS coefficients seem to be inconsistent. We therefore resort to the IV coefficients to document the negative and significant impact of aid on PFIQ in the sample without outliers. The data necessary to compute the PFIQ score is scarce across certain countries.
Specifically, government expenditure and tax revenue were poorly sourced. The unavailability of data on tax revenue could be indicative of a poorly performing public finance system. The concern was then that running two regressions, one including the countries for which data was scarce, the other excluding those countries, would lead to differing results. However, our concern was alleviated by the fact that, excluding outliers, samples with complete- and incomplete data countries yield overall the same results (Table 3, Eqs. 9, 10, 11 and 12). These results greatly reduce our concerns about sample selection bias.
For the sake of brevity, in this section we directly report the main results using the balanced sample, excluding outliers. The complete set of regression results (analogous to the PFIQ regressions) can be found in Table 4. We mainly rely on the results from the regressions run in the sample excluding outliers, given the distortions caused by outliers in PFIQ-regressions.
Our findings from the PFIQ equation excluding outliers show aid instrumented for has a significant negative effect on PFIQ. We cannot therefore disregard the need to instrument for PFIQ, given these results. For both Eqs. 7 and 8 (Table 4), we use military in politics, internal conflict and external conflict as instruments for the PFIQ variable. These two regressions yield similar results. In either case, the PFIQ- coefficient is significant and positive. All other control variables remain significant, except measure of openness and democracy dummy; and, explanatory power is acceptable (respectively 42% and 36% for Eqs. 7 and 8).
Conducting the Durbin-Wu-Hausman test for systematic difference in the OLS and IV coefficients for regression pairs 5-7 and 6-8, values of the test- statistics do not allow us to reject the null hypothesis of no difference between OLS and IV coefficients at the 10% level. We are therefore able to use OLS coefficients, for matters of efficiency, in order to document a significant positive impact of PFIQ onto aid over GDP and onto aid per capita.
Another interesting result we would like to highlight is that the democracy dummy is insignificant in all specifications 5-8. This gives support to our initial assumption that multilateral aid is not primarily directed by political considerations.
Comparing coefficients across Eqs. 5-8 and 9-12 (Table 4), results are of comforting similarity. The PFIQ coefficient has a significant and positive impact on aid-flows in all cases; the only difference in control variables is that openness becomes significant in Eq. 11.
Perhaps the most serious limitation to our findings is the possibility of reverse causality embedded in our model specifications. Nonetheless, provided the variables used for 2SLS estimation in our two main equations are acceptably exogenous, the regressions do yield strong results. Specifically, reverse causality seems to not affect OLS estimates of the impact of aid onto PFIQ.
It must be noted that the data set was far from being complete. Perhaps missing data on certain variables which enter the computation of the PFIQ index are already indicative of a poorly-performing public finance system; or, perhaps missing data reflect negligence on behalf of policy-makers to seize the importance of public finance in implementing successful and sustainable policies in developing countries.
It emerged from discussions with policy makers and other experts that there could be region-specific characteristics which affect public finance quality. We intend to alleviate this potential unobserved heterogeneity using either fixed effects, or first differences estimation techniques. Furthermore, we realized ex post that we have as yet not conducted the F-statistic benchmark test in the first stage of our 2SLS approach.
There are three further working hypotheses we would like to test. Firstly, we are interested in whether one can find bi-lateral donor countries (which are not considered in the present study), whose aid-flow patterns resemble those of the multilateral donors we have investigated. Secondly, we remain inclined to link our investigations with real-world policy analysis, by testing whether the originators of the PEFA framework (2) do respond to changes in our PFIQ Index. Thirdly, one of our initial hypotheses was that for extremely poor countries donors give aid regardless of the recipient country's institutional performance ("every dollar counts"), whereas institutional and policy factors may carry more weight in aid disbursements to relatively more developed countries. We would like to test this hypothesis by investigating potential differences in the impact of PFIQ changes on multilateral aid receipts for country groups of different income levels.
Our empirical findings help to shed light on many of the puzzles and questions discussed in our literature review and working hypotheses. The general picture is that multilateral aid-flows have a negative impact on a recipient country's Quality of Public Financial Institutions (PFIQ), and that better PFIQ has a positive effect on a given country's multilateral aid receipts.
From our literature review, existing studies show that aid mostly has a negative impact on recipients' institutions and governance, including corruption, democratization, tax effort and more. The findings of our paper now suggest that the same causal direction applies to the relationship between multilateral aid and PFIQ.
In the same section we also saw that, empirically, aid does have a positive effect on the size of receiving governments, whilst these changes fail to translate into economic growth and poverty alleviation. We consider our findings to be helpful insights into the black box of governance. Its failure to turn aid receipts into desirable results seems partly attributable to the fact that multilateral aid, under its current form, is not suited to improve a given country's public finance quality. This arguably is a missing link between aid-flows and improved economic outcomes.
The second part of our literature review did not allow us to draw any clear-cut assumptions about the effect of PFIQ on multilateral aid. For our panel, at least, we are now confident to conclude that exogenous improvements in a recipient country's Quality of Public Financial Institutions attract increased aid-flows from multilateral donors. This suggests that, while multilateral aid seems to fail to have a positive impact on PFIQ, international donor organizations reward exogenous improvements in the quality and reliability of the public financial systems in target countries.
Appendix (full version available on request)
Acknowledgements We are grateful for all support received from the LSE Economics Department. In particular, we are heavily indebted to Dr. Judith Shapiro, our undergraduate tutor, and our fellow participants at the LSE Economics Undergraduate Research Workshop. This project would not have been nearly as productive and enjoyable without the numerous discussions with our critical and helpful friends, both in the Research Workshops, and across different academic departments. The instructive comments of fellow participants at the Carroll Round Conference at Georgetown University, as well as at the International Atlantic Economic Society's Conference, helped in finalising the study. We also thank the BLPES' Data Library for helpful support. For any mistakes or inconsistencies, the authors blame each other (This work is the outcome of a joint effort by the authors while they were undergraduate students.).
Published online: 11 April 2010
Afonso, A., Ebert, W., Schuknecht, L., & Thone, M. (2005). Quality of Public Finances and Growth. European Central Bank Working Paper Series: No.438; February Issue.
Alesina, A., & Dollar, D. (2000). Who gives foreign aid to whom and why? Journal of Economic Growth, 5(1), 33453.
Alesina, A., & Weder, B. (2002). Do corrupt governments receive less foreign aid? American Economic Review, 92(4), 1126-1137.
Boone, P. (1996). Politics and the effectiveness of foreign aid. European Economic Review, 40(2), 289-329.
Brautigam, A. D., & Knack, S. (2004). Foreign aid, institutions and governance in Sub-Saharan Africa. Economic Development and Cultural Change, 52(2), 255-285.
Burnside, C., & Dollar, D. (2000). Aid, policies and growth. American Economic Review, 90(4), 847-868.
Easterly, W., Levine, R., & Roodman, D. (2004). Aid, policies, and growth: comment. American Economic Review, 94(3), 774-780.
King, R. G., & Levine, R. (1993). Financial intermediation and economic development. In C. Mayer & X. Vives (Eds.), Financial intermediation in the construction of Europe (pp. 156- 189). London: Centre for Economic Policy Research.
Knack, S. (2001). Aid dependence and the quality of governance. Southern Economic Journal, 68(2), 310-329.
Organization for Economic Cooperation and Development. (2006). 2006 Survey on Monitoring the Paris Declaration. OECD Journal of Development, 8(2), 1-131.
Public Expenditure and Financial Accountability (2005). public financial management--performance measurement framework. Unpublished working paper, PEFA Secretariat, World Bank.
Van Rijckeghem, C., & Weder, B. (2001). Bureaucratic corruption and the rate of temptation: do wages in the civil service affect corruption, and by how much? Journal of Development Economies, 65(2), 307-31.
M. Gstoettner * A. Jensen ([mail])
London School of Economics and Political Science, London, UK
(1) Survey on Monitoring the Paris Declaration (2006), OECD, p.11
(2) The World Bank, the European Commission (EC), the UK's Department for International Development (DFID), the Swiss State Secretariat for Economic Affairs (SECO), the Royal Norwegian Ministry of Foreign Affairs, the French Ministry of Foreign Affairs, and the International Monetary Fund (IMF).
Table 5 Country-specific summary statistics (86 countries) Country Mean(gdppc) Mean(dem) Mean(aidP) Albania 1017.047 .7 4.31e-08 Algeria 1816.059 0 1.43e-09 Argentina 6934.738 1 1.56e-10 Armenia 594.8647 .9 7.14e-08 Azerbaijan 784.2749 .4 2.19e-08 Bahrain 10851.42 0 5.16e-10 Bangladesh 289.3903 .65 2.94e-08 Belarus 1219.093 .6 Bolivia 929.7963 1 3.58e-08 Botswana 2717.025 1 1.23e-08 Brazil 3538.494 .95 9.75e-11 Burkina Faso 195.3379 .15 9.44e-08 Cameroon 706.6931 0 1.82e-08 Chile 3847.072 .75 1.58e-10 China 626.2484 0 1.46e-09 Colombia 1920.82 1 3.60e-10 Congo 1111.762 .25 1.03e-08 Congo, DR 152.678 0 4.21e-08 Costa Rica 3458.983 1 2.03e-09 Croatia 3882.315 .55 1.79e-09 Cyprus 11201.71 1 3.85e-09 Dominican Republic 1794.643 1 3.21e-09 Ecuador 1323.028 1 4.40e-09 Egypt 1277.502 0 3.32e-09 El Salvador 1857.172 1 6.17e-09 Ethiopia 123.9293 .45 9.15e-08 Gabon 4635.22 0 3.56e-09 Gambia 305.7655 .5 1.52e-07 Ghana 223.0734 .4 1.05e-07 Guatemala 1552.921 .9 3.73e-09 Guinea 353.4613 0 7.97e-08 Guinea-Bissau 170.2102 .35 3.04e-07 Guyana 799.8542 .6 1.24e-07 Haiti 534.7904 .3 2.62e-08 Honduras 1098.772 1 2.98e-08 Indonesia 700.5228 .25 1.27e-09 Iran 1448.69 .35 5.30e-10 Israel 16261.46 1 1.09e-10 Jamaica 2889.925 1 4.96e-09 Jordan 1800.009 0 1.92e-08 Kazakstan 1277.624 .35 1.53e-09 Kenya 418.5943 .1 2.82e-08 Kuwait 16011.9 0 6.56e-11 Lebanon 4080.84 0 5.89e-09 Liberia 254.903 .3 2.22e-07 Malawi 140.6572 .5 2.10e-07 Malaysia 3058.58 1 2.95e-10 Mali 226.0282 .6 1.19e-07 Mexico 5237.651 .8 5.58e-11 Mongolia 466.1763 .7 5.48e-08 Morocco 1235.114 0 5.77e-09 Myanmar 0 Namibia 1751.041 1 1.79e-08 Nicaragua 744.5935 .7 5.46e-08 Nigeria 358.0599 .25 2.91e-09 Oman 7597.145 0 3.20e-10 Pakistan 486.0771 .55 1.46e-08 Panama 3468.971 .75 7.38e-10 Papua New Guinea 666.9385 1 2.00e-08 Paraguay 1386.417 .75 4.01e-09 Peru 1974.254 .9 1.29e-09 Philippines 921.5164 .85 1.92e-09 Qatar 0 Senegal 443.676 .2 6.39e-08 Sierra Leone 212.8899 .15 1.20e-07 Singapore 17290.8 0 5.47e-11 Slovenia 8754.697 1 3.95e-09 South Africa 3088.067 .9 1.08e-09 Sri Lanka 313.7842 1 2.55e-08 Sudan 316.0038 .15 4.16e-08 Syria 532.5496 0 4.37e-09 Tanzania 263.1885 .2 5.53e-08 Thailand 3290.235 .95 9.59e-10 Togo 255.6777 0 7.30e-08 Trinidad & Tobaco 5563.097 1 1.71e-09 Tunisia 1700.064 0 8.17e-09 Turkey 3472.781 1 3.79e-10 Uganda 198.2161 .05 1.04e-07 Ukraine 931.9698 1 United Arab Emirates 24077 0 3.84e-11 Uruguay 5452.082 .95 4.19e-10 Venezuela 4924.516 1 7.33e-11 Vietnam 298.514 0 1.28e-08 Yemen 484.6299 .3 2.00e-08 Zambia 348.7018 .65 1.15e-07 Zimbabwe 604.5765 .15 1.49e-08 Country Mean(aidPC) Mean(PFIQ) Albania .0000419 1.416109 Algeria 2.57e-06 1.430711 Argentina 1.09e-06 1.817193 Armenia .0000403 1.210326 Azerbaijan .0000133 Bahrain 5.53e-06 Bangladesh 7.98e-06 .8521771 Belarus 1.641334 Bolivia .000033 .9341965 Botswana .0000271 2.068361 Brazil 3.44e-07 1.774034 Burkina Faso .0000183 1.54116 Cameroon .000012 1.568563 Chile 7.38e-07 1.926549 China 6.98e-07 1.607048 Colombia 6.43e-07 1.280493 Congo .0000115 Congo, DR 5.87e-06 49.5945 Costa Rica 5.70e-06 1.927537 Croatia 6.99e-06 2.289571 Cyprus .0000368 Dominican Republic 5.32e-06 1.566392 Ecuador 5.73e-06 226.4259 Egypt 4.11e-06 1.483343 El Salvador .0000109 .9745212 Ethiopia .0000112 1.081786 Gabon .0000163 1.441774 Gambia .000047 1.43724 Ghana .0000228 1.304958 Guatemala 5.71e-06 .9562021 Guinea .0000278 1.493075 Guinea-Bissau .0000522 .6470916 Guyana .0000963 .2673844 Haiti .0000135 Honduras .0000328 .5639676 Indonesia 7.82e-07 1.071712 Iran 7.44e-07 1.797648 Israel 1.69e-06 2.653183 Jamaica .0000142 Jordan .000033 1.805811 Kazakstan 1.77e-06 2.016615 Kenya .000012 1.818444 Kuwait 1.29e-06 2.001982 Lebanon .0000223 1.81491 Liberia .0000278 Malawi .0000293 1.244501 Malaysia 7.41e-07 2.296094 Mali .0000263 .4954835 Mexico 2.94e-07 1.849994 Mongolia .0000242 Morocco 7.24e-06 1.821694 Myanmar 2.14e-06 Namibia .000031 2.526974 Nicaragua .00004 1.073298 Nigeria 1.06e-06 Oman 2.27e-06 1.975038 Pakistan 6.90e-06 1.03514 Panama 2.41e.06 1.14213 Papua New Guinea .000013 1.996102 Paraguay 5.57e-06 1.070044 Peru 2.53e-06 1.054426 Philippines 1.75e-06 1.227474 Qatar 2.53e-06 Senegal .0000284 1.377543 Sierra Leone .0000236 Singapore 7.22e-07 Slovenia .0000366 2.65573 South Africa 3.27e-06 2.012196 Sri Lanka 7.50e-06 1.624376 Sudan .0000118 .6274182 Syria 2.27e-06 1.2091 Tanzania .0000142 Thailand 2.61e-06 1.949288 Togo .0000193 .9950033 Trinidad & Tobaco 8.53e-06 1.968389 Tunisia .0000139 1.898433 Turkey 1.34e-06 1.309862 Uganda .0000197 1.13695 Ukraine 1.828316 United Arab Emirates 9.79e-07 1.963742 Uruguay 2.25e-06 1.536148 Venezuela 3.54e-07 1.487722 Vietnam 4.34e-06 Yemen .0000122 1.460926 Zambia .0000383 1.366288 Zimbabwe 9.04e-06 1.28862 Table Country, contents(mean gdppc mean dem mean aidP mean aidPC mean PFIQ) Table 1 Data summary--all observations STATA Variable Name Formula Observations Abbreviation aid Aid -- 1,573 aidP Aid over GDP aid/gdp 1,525 aidPC Aid Per Capita aid/pop 1,570 bur Bureaucracy Quality -- 1,590 burp Bureaucracy Quality bur/4 1,590 Percent corr Corruption -- 1,591 corrP Corruption Percent corr/6 1,591 Country Country -- 86 countries dem Democracy Dummy =1 if demindex 1,720 >=0; =0 otherwise demindex Democracy Index -- 1,659 exp Exports as % of GDP -from dataset 1,597 extcon External Conflicts -- 1,590 gdp GDP -- 1,626 gdppc GDP Per Capita gdp/pop 1,623 govexp Government -- 1,320 Expenditure illit Illiteracy (100--lit) 1,422 imp Imports as % of GDP -from dataset 1,596 infl Inflation -- 1,455 intcon Internal Conflicts -- 1,590 laword Law and Order -- 1,590 lawordP Law and Order Percent laword/6 1,590 lit Literacy -- 1,422 loggdppc Log of GDP Per Capita ln(gdppc) 1,623 logpop Log of Population ln(pop) 1,717 M2 Broad Money -compiled in 1,554 dataset mil Military in Politics -- 1,586 mort Infant Mortality -- 1,720 openness Economic Openness imp+exp 1,596 PFIQ Quality of Public burP+lawordP 870 Financial Institutions +corrP+taxe ff+surdef PFIQClean Quality of Public burP+lawordP 856 Financial Institutions +corrP+taxe w/o Outliers ff+surdef pop Population -- 1,717 surdef Government Surplus/ (taxrev-- 900 Deficit govexp)/gdp taxeff Tax Effort taxrev/gdp 1,104 taxrev Tax Revenue -- 1,156 volfin Volume of Financial M2/gdp 1,505 System Year Year -- 1,720 STATA Source Comments Abbreviation aid OECD DAC aidP -- aidPC -- bur ICRG burp -- corr ICRG corrP -- Country -- dem -- demindex Polity IV Index exp WB World Development Indicators extcon ICRG gdp WB World Development Indicators gdppc -- govexp WB World Development Indicators illit -- imp WB World Development Indicators infl WB World Development Indicators intcon ICRG laword ICRG lawordP -- lit UNESCO Database loggdppc -- logpop -- M2 IMF Computed by hand, International summing lines 14A..ZF, Financial 24...ZF and 25...ZF Statistics mil ICRG mort WB World Development Indicators openness -- PFIQ -- PFIQClean -- PFIQClean==PFIQ if PFIQ<4 pop WB World Development Indicators surdef -- taxeff -- taxrev IMF Historical values (before Government 1990) collected country Finance by country Statistics volfin -- Year -- Table 2 Summary of regression specification and identification Variable Equation Variants of (1), PFIQ Possibly endogenous variables PFIQ LHS LHS PFIQClean LHS LHS Aid/GDP RHS RHS Aid/Population RHS RHS Exogenous variables Logarithm of GDP/ Included Included Included Included Population Logarithm of population Volume of financial Included Included Included Included system Military in Politics Included Included Included Included Internal Conflict Included Included Included Included External Conflict Included Included Included Included Inflation rate Included Included Included Included Openness Mortality Illiteracy Democracy dummy Variable Equation Variants of (2), Aid Possibly endogenous variables PFIQ RHS RHS PFIQClean RHS RHS Aid/GDP LHS LHS Aid/Population LHS LHS Exogenous variables Logarithm of GDP/ Included Included Included Included Population Logarithm of Included Included Included Included population Volume of financial system Military in Politics Internal Conflict External Conflict Inflation rate Openness Included Included Included Included Mortality Included Included Included Included Illiteracy Included Included Included Included Democracy dummy Included Included Included Included LHS indicates that a variable is included as the left-hand side variable. RHS indicates that the variable is included as a right-hand side variable. All exogenous variables which serve as control variables are potentially used as instruments in 2SLS estimation Table 3 PFIQ regressions summary (t-statistics in parentheses) Dependent Variable: PFIQ Sample: All Observations Estimation Method: OLS Equation: 1 2 Aid Variable: Aid/GDP Aid/ Population Constant 13.31 (0.96) 7.94 (0.67) Aid -4.92e+07 -115060.3 (-1.26) (-1.11) Log GDP per capita -1.65 (-0.81) -0.81 (-0.47) Volume of Financial -1,708,544 -1,706,065 System (-0.54) (-0.54) Military in Politics (a) -0.75 (-0.61) -0.76 (-0.65) External Conflicts (a) -0.77 (-1.04) -0.83 (-1.11) Internal Conflict (a) 1.74 (2.30) 1.78 (2.34) Inflation Rate 0.03 (1.17) 0.03 (l.09) Number of Observations 711 711 Adjusted R^2 0.00 0.00 Durbin-Wu-Hausman -- -- Test Statistic Dependent Variable: PFIQ Sample: All Observations Estimation Method: 2SLS Equation: 3 4 Aid Variable: Aid/GDP Aid/ Population Constant 17.10 (0.49) -4.04 (-0.26) Aid -5.28E+07 130715.9 (-0.34) (0.59) Log GDP per capita -1.70 (-0.34) 0.64 (0.31) Volume of Financial -2,993,198 -923771.1 System (-0.63) (-0.29) Military in Politics (a) -1.02 (-0.59) -1.27 (-1.03) External Conflicts (a) -1.13 (-1.24) -0.57 (-0.73) Internal Conflict (a) 1.96 (2.27) 1.53 (1.93) Inflation Rate 0.03 (1.21) 0.03 (1.15) Number of Observations 609 711 Adjusted R^2 0.00 0.00 Durbin-Wu-Hausman 0.15 1.58 Test Statistic Dependent Variable: PFIQClean Sample: W/o PFIQ Outliers Estimation Method: OLS Equation: 5 6 Aid Variable: Aid/GDP Aid/ Population Constant 0.18 (1.37) 0.14 (1.25) Aid -614987.5 -2046.76 (-1.69) (-2.13) Log GDP per capita 0.02 (0.91) 0.02 (1.49) Volume of Financial 78220.81 76330.59 System (2.69) (2.62) Military in Politics (a) 0.17 (15.63) 0.17 (15.74) External Conflicts (a) 0.02 (2.35) 0.01 (2.13) Internal Conflict (a) 0.07 (10.10) 0.07 (10.21) Inflation Rate 0 (-3.92) 0 (-4.06) Number of Observations 699 699 Adjusted R^2 0.58 0.58 Durbin-Wu-Hausman -- -- Test Statistic Dependent Variable: PFIQClean Sample: W/o PFIQ Outliers Estimation Method: 2SLS Equation: 7 8 Aid Variable: Aid/GDP Aid/ Population Constant 2.59 (6.10) 0.46 (3.13) Aid -1.14e+07 -8497.98 (-6.02) (-4.06) Log GDP per capita -0.32 (-5.27) -0.01 (-0.73) Volume of Financial 31151.83 55600.56 System (0.54) (1.82) Military in Politics (a) 0.24 (11.24) 0.18 (15.58) External Conflicts (a) 0.02 (1.33) 0.01 (1.30) Internal Conflict (a) 0.07 (6.73) 0.08 (10.47) Inflation Rate 0 (-1.59) 0 (-3.99) Number of Observations 597 699 Adjusted R^2 0.09 0.56 Durbin-Wu-Hausman 41.4 12.05 Test Statistic Dependent Variable: PFIQClean Sample: Reduced Dataset w/o PFIQ Outliers (b) Estimation Method: OLS Equation: 9 10 Aid Variable: Aid/GDP Aid/ Population Constant 0.28 (2.09) 0.31 (2.59) Aid -346405.4 -1990.02 (-0.87) (1.85) Log GDP per capita 0 (0.18) 0 0 Volume of Financial 77265.05 72889.63 System (2.64) (2.48) Military in Politics (a) 0.17 (15.26) 0.18 (15.37) External Conflicts (a) 0.01 (2.19) 0.02 (2.06) Internal Conflict (a) 0.07 (9.13) 0.07 (9.29) Inflation Rate 0(-3.97) 0(-4.06) Number of Observations 637 637 Adjusted R^2 0.56 0.56 Durbin-Wu-Hausman -- -- Test Statistic Dependent Variable: PFIQClean Sample: Reduced Dataset w/o PFIQ Outliers (b) Estimation Method: 2SLS Equation: 11 12 Aid Variable: Aid/GDP Aid/ Population Constant 3.22 (5.09) 0.62 (3.77) Aid -1.52e+07 -8078.82 (-4.68) (-3.35) Log GDP per capita -0.41 (-4.56) -0.04 (-1.73) Volume of Financial 15198.78 52210.33 System (0.22) (1.69) Military in Politics (a) 0.24 (9.52) 0.19 (15.03) External Conflicts (a) 0.02 (1.26) 0.01 (1.35) Internal Conflict (a) 0.08 (5.82) 0.08 (9.49) Inflation Rate 0(-0.98) 0(-3.99) Number of Observations 547 637 Adjusted R^2 Not reported 0.54 Durbin-Wu-Hausman Not reported 33.5 Test Statistic (a) NB: for these variables higher ICRG scores indicate less military involvement and fewer conflicts, respectively (b) NB: the reduced dataset consists of those 48 countries (out of 86) for which the relevant data were almost complete (see Appendix Tables II. & III.) In Eqs. 3, 7 and 11 the exogenous variables for 2SLS estimation are: Illiteracy and Infant Mortality as Instruments for Aid (i.e. Aid/GDP) In Eqs. 4, 8 and 12 the exogenous variables for 2SLS estimation are: we use Log of Population to instrument for Aid (i.e. Aid/Population) Table 4 Aid regressions summary (t-statistics in parentheses) Dependent Variable: Aid/GDP Aid/ Aid/ Aid/ Population GDP Population Sample: All Observations Estimation Method: OLS 2SLS (a) Equation: 1 2 3 4 PFIQ Variable: PFIQ Constant 2.79e-07 0 (11.49) -- -- (7.51) PFIQ Score -8.33e-11 2.66e-08 -- -- (-2.52) (-2.43) Log of GDP per capita -1.68e-08 -5.75e-06 -- -- (-6.85) (-7.04) Log of Population -9.98e-09 -5.64e-06 -- -- (-7.82) (-13.30) Economic Openness 7.72e-13 1.00e-08 -- -- (0.01) (0.57) Democracy Dummy 1.07e-09 1.66e-06 -- -- (0.30) (1.39) Infant Mortality 7.37e-10 1.06e-7 (3.12) -- -- (7.22) Illiteracy -3.37e-10 -1.15e-07 -- -- (-2.70) (-2.76) Number of Observations 709 709 -- -- Adjusted R^2 0.41 0.35 -- -- Durbin-Wu-Hausman Test -- -- -- -- Statistic Dependent Variable: Aid/GDP Aid/ Aid/GDP Aid/ Population Population Sample: W/o PFIQ Outliers Estimation Method: OLS 2SLS Equation: 5 6 7 8 PFIQ Variable: PFIQClean Constant 2.75e-07 0 (11.47) 2.73e-07 0(11.26) (7.44) (7.32) PFIQ Score 9.93e-09 4.04e-06 1.93e-08 6.33e-06 (3.34) (4.09) (4.61) (4.51) Log of GDP per capita -1.78e-08 -6.13e-11 -1.86e-08 -6.23e-06 (-7.24) (-7.51) (-7.43) (-7.55) Log of Population -1.02e-08 -5.75e-06 -1.06e-08 -5.80e-06 (-8.00) (-13.59) (-8.16) (-13.59) Economic Openness -3.92e-11 -7.68e-09 -8.76e-11 -1.88e-08 (-0.71) (-0.42) (-1.52) (-0.99) Democracy Dummy 9.22e-11 1.37e-06 -2.55e-10 1.82e-06 (0.03) (1.15) (-0.07) (1.07) Infant Mortality 8.52e-10 1.42e07 9.09e-10 1.57e-07 (8.19) (4.12) (8.55) (4.48) Illiteracy -4.63e-10 -1.53e-07 -5.15e-10 -1.63e-07 (-3.66) (-3.64) (-3.99) (-3.83) Number of Observations 695 695 694 694 Adjusted R^2 0.43 0.37 0.42 0.36 Durbin-Wu-Hausman Test -- -- 10.18 7.18 Statistic Dependent Variable: Aid/GDP Aid/ Aid/GDP Aid/ Population Population Sample: Reduced Dataset w/o PFIQ Outliers (b) Estimation Method: OLS 2SLS Equation: 9 10 11 12 PFIQ Variable: PFIQClean Constant 2.37e-07 0(9.42) 2.38e-07 0(9.40) (7.52) (7.51) PFIQ Score 9.26e-09 3.99e-06 1.48e-08 6.46e-06 (3.96) (4.13) (4.50) (4.74) Log of GDP per capita -1.84e-08 -5.82e-06 -1.89e-08 -6.03e-06 (-9.22) (-7.04) (-9.37) (-7.23) Log of Population -6.92e-09 -4.96e-06 -7.21e-09 -5.09e-06 (-6.32) (-10.94) (-6.51) (-11.10) Economic Openness -8.22e-11 7.83e-09 -1.10e-10 -4.60e-09 (-1.63) (0.37) (-2.12) (-0.21) Democracy Dummy 5.76e-09 1.75e-06 5.48e-09 1.63e-06 (2.01) (1.48) (1.90) (1.36) Infant Mortality 3.87e-10 1.35e-07 4.21e-10 1.50e-07 (4.46) (3.76) (4.77) (4.10) Illiteracy -1.57e-10 -1.12e-07 -1.90e-10 -1.27e-07 (-1.53) (-2.66) (-1.84) (-2.97) Number of Observations 636 636 636 636 Adjusted R^2 0.37 0.33 0.36 0.32 Durbin-Wu-Hausman Test -- -- 5.76 6.59 Statistic (a) NB: In Section VIII. It was found that, for the unbiased sample, exogenous aid-shocks do not have an impact on PFIQ, consequently we do not instrument for PFIQ in the present specification6 NB: the reduced dataset consists of those 48 countries (out of 86) for which the relevant data were almost complete (see Appendix Tables II. & III.) In Eqs. 7, 8, 11 and 12 the exogenous variables for 2SLS estimation are: Military in Government, External Conflicts and Internal Conflicts
|Printer friendly Cite/link Email Feedback|
|Author:||Gstoettner, Markus; Jensen, Anders|
|Publication:||Atlantic Economic Journal|
|Date:||Jun 1, 2010|
|Previous Article:||Pollution, factor ownerships, and emission taxes.|
|Next Article:||The impact of Microeconomics: The Puzzle Game on student evaluations.|