Institutional Quality, EU-skepticism and Regional Funds in Central & Eastern Europe
X.B. Steijlen x.b.steijlen@student.rug.nl
MSc Economics Thesis (EBM877A20) University of Groningen
Supervisor: dr. C.G.F. (Christiaan) van der Kwaak
Abstract: This paper analyzes the effect of European Structural and Investment Funds on economic growth in the European member states and the CEE countries in particular. The main analysis is on the effect of these funds interacting with institutional quality at different time lags. I find no robust result that institutional quality affects the effectiveness of ESI funds. The results do point out that ESI funds have a positive effect on growth at one-, two-, and three- year lags. This result at the one-year lag is robust to a new instrument proposed by this paper: EU skepticism in a country’s electorate.
2 I. Introduction
Over its entire existence, the EU has devoted a significant part of its budget towards Economic and Regional Development.1 Some say, transferring those
sums to those goals is an outdated policy. For example, Prime Minister of the Netherlands, Mark Rutte, outlined this in his speech on the future of Europe in Berlin.2 He spoke how he is hesitant to pay more for a larger EU budget. He does not want to add the costs of the challenges of today (climate change, migration control, common external borders, collective security) to the costs of the policies created in the ‘90s (agricultural funds, cohesion & development funds). The imminent departure of the UK also means a significant net-payer will no longer contribute to the budget. It is reasonable to argue that fewer net-paying members should result in a lower budget, not a higher one. It is just as reasonable to argue that the EU faces new challenges today and therefore actually does require a larger budget. In any case, rising populism across the member states makes it harder for governments to justify an increase in the EU budget (Inglehart and Norris (2016)). Therefore it is relevant at this point to research how effective the spending on the established priorities has been. Do European Regional Development Funds (ERDF) and European Social Funds (ESF) contribute towards economic growth in proportion to their size? Have they actually been successful in promoting its main goal: the growth of the economy in less developed regions? If it turns out that they have largely been unsuccessful, this offers interesting insights for policymakers who struggle with the budgetary question. Say that ESI funds offer no significant or only a small benefit, they may than decide to devote a lesser part of the total budget to structural policy in the next multiannual financial framework and focus on more recent challenges.
1 Throughout the paper I will use the terms regional funds, ESI funds and structural funds
interchangeably. Note that in all cases I mean all funds falling under the ERDF and ESF program.
2 Rutte, Mark. “Underpromise and Overdeliver: fulfilling the promise of Europe.” Speech, Bertelsmann
3 Furthermore, accession negotiations are being held with the former Yugoslav Republic of Macedonia, Montenegro, Serbia and Albania. If these countries were to join they will add more strain on the ESI funds, seeing as how they will most likely be net-receiving countries of the budget. These countries’ democratic and economic institutions, although passing the EU ‘exam’, will not be of equal quality as their future co-members in the West. By identifying whether ESI funds are successful and under which institutional circumstances, may prove a worthwhile exercise both for the EU as a whole and the future growth of the new member states in particular. The European Commission can then decide which types of institutions require improvement first, in order to maximize the return (economic growth) on investment of the funds.
For the abovementioned reasons, I will test the following hypotheses. First and foremost, what is the effect of ESI funds on growth? If such an effect exists, does interacting with institutional quality indicators amplify/mitigate the positive/negative effect? Third, after how many years will the effects of ESI funds become apparent? Fourth, do these institutional indicators matter more in CEE countries than in the Western European countries? In the next section, I will review the literature. Subsequently, I describe the research method used to test these hypotheses. In the fourth section, I will discuss the results. In the last section, I will conclude.
II. Literature Review
Multiple papers have assessed the effect of the ESI funds on economic growth; Ederveen, de Groot, and Nahuis (2006, henceforth EGN), Cappelen et. al (2003), Mohl and Hagen (2010), Dall’erba and Le Gallo (2008). Their approaches mainly differ in regional vs. country level analyses and measuring the effects of ERDF individually or adding interaction effects with institutional quality.
4 period of 1960 to 2005. They opt for estimation at the country-level instead of the regional level for two reasons: They argue that at the country-level the estimation will be less sensitive to leakage or spillover effects. Furthermore, EGN want to control for institutional quality and education attainment. Data which is not available at the regional-level. Finally, they argue that regional analysis suffers from endogeneity. Reverse causation exists due to EU funds influencing GDP, while GDP in turn influences which regions receive funds. According to EGN, this is much less of a problem at the country level, because each country has (relatively) rich as well as (relatively) poor regions and that cross-border regional spillovers are of a small magnitude. Relatively poor regions are regions that receive relatively few funds and relatively rich regions receive a relatively high amount of funds. Their analysis finds that structural funds do not influence differences in GDP growth among EU states. However, they do find that structural funds lead to increased GDP growth in countries with a high level of institutional quality.
Cappelen et. al (2003) estimate the effectiveness of structural funds on the regional level, not interacting with institutions such as corruption or openness. Instead, they identify three development factors: diffusion of knowledge, innovation, complementary factors (infrastructure, Population Density, unemployment). Their estimation yields that structural funds have a positive impact across the board, but that their positive impact is most striking in developed economies.
5 bias is often overlooked in the previous literature with the exception of Dall’erba and Le Gallo (2008), who use three different instruments to solve the endogeneity issue, among which are; the distance by road to Brussels and the travel time from the most populated town of each region to Brussels. The idea is that the distance-to-Brussels is independent of GDP growth, while structural funds follow a center-periphery distribution and are therefore affected by the distance-to-Brussels. Mohl and Hagen (2010) take a different approach to mitigating the endogeneity bias by using a two-step GMM estimator with lagged variables. In line with Becker, Egger and von Ehrlich (2010), they find that the sum of Objective 1 2 and 3 payments show no significant impact on growth. However, Objective 1 payments individually with a time lag up to four years do show significant and robust positive results at the regional level when accounting for spillovers. They assume this is due to Objective 2 and 3 projects not being based on clear cut objectives. Which allows for political bargaining that results in projects that are based on political outcomes instead of economic outcomes. They find that the effect of ESI funds changes in sign when taking different lags of the variable into account.
6 Bräutigam (2000) shows how, in the absence of proper monitoring, the inflow of foreign money incentivizes rent-seeking. This is also argued by EGN. It will dampen or even eliminate any positive effect on GDP growth. Also, because other resources such as human capital will be redirected to inefficient exercises such as rent-seeking. Apart from corruption, it is also important to look at the competence of governments. All regional funds must be co-financed by the country’s government. They must be able to plan such costs in their budget even if regional policy is not on the policy agenda. In their seminal paper Sachs et. al (1995) outline how useful and accurate it is to measure a country’s economic reform policies by the country’s degree of trade liberalization. They argue that once a country opens up to international competition it is forced to reform other parts of the economy as well, in order to improve the competitiveness of domestic businesses. In other words, the progress toward trade liberalization goes hand in hand with developing ‘good’ institutions. Furthermore Acemoglu et. al (2003) and Easterly (2005) describe how “stable policies” contribute to economic growth. In the paper of Acemoglu et. al (2003) “stable policies” are low inflation and balanced budgets. They outline how countries that pursued high inflation and unbalanced budgets realized lower economic growth. Easterly (2005) shows how high inflation and large budget deficits are the result of bad policies.
The current literature lacks research on the effect of structural funds that includes all EU countries of the 2004 accession (EU-10).3 Furthermore, no research has been done on the most recent accurate data: the time period of 2006-2013. My paper will fill this gap. It is important to analyze the 2004 countries because it are those countries which are most similar to the Balkan EU-candidates. EGN conclude that the EU regional fund policy should be redesigned: Funds should first be focused on institution building, so they can be effective after institutions are of proper quality. However, the 2006-2013
7 multiannual financial framework did not identify institution-building as a requirement. EGN’s policy recommendation has not been implemented. The empirical evidence does show that the CEE countries are rapidly catching up with the rest of Europe in terms of GDP per capita, Rapacki and Próchniak (2009). I will analyze whether institutional quality matters for the effect of ESI funds in CEE countries.
8 III. Methodology
In line with the literature (EGN) the research questions will be tested using the neoclassical growth model as proposed by Mankiw, Romer and Weil (1992). The regression features an augmented Solow-Swan model (Solow (1956), Swan(1956)) which includes the savings rate 𝑠𝑖𝑡, human capital, population growth 𝑛𝑖𝑡, technological progress 𝑔𝑎 and the depreciation rate 𝛿. The
dependent variable 𝑔𝑖𝑡 is defined as the GDP per capita growth rate annually in percentages, taken from the World Bank Development Indicators. I expect GDP per capita growth in the previous year to influence the economic growth in the current year, therefore I include a one-year lag of the dependent variable. Human capital will be proxied by secondary education attainment (a common proxy for human capital used widely in the economic growth literature). It is however quite a broad measure and must, therefore, be taken with a grain of salt. The data is taken from Eurostat and details the upper-secondary educational attainment for the ages 25-64 as a percentage of the total amount of people in that age range. The gross domestic savings rate as a percentage of GDP is taken from the World Bank Development Indicators. As well as the annual percentage change in population 𝑛𝑖𝑡. In line with the literature, 𝛿 + 𝑔𝑎 will be set to 5
percent for all countries. Subscripts ‘it’ indicate the country and year respectively, while subscript ‘-p’ indicates a lag of year(s) p.
(1) 𝑔𝑖𝑡 = 𝛽0+ 𝛽1𝑔𝑖𝑡−1+ 𝛽2𝑠𝑖𝑡 + 𝛽3ℎ𝑢𝑚𝑖𝑡+ 𝛽4(𝑛𝑖𝑡+ 𝛿 + 𝑔𝑎) +
𝛽5𝐸𝑆𝐼𝐹𝑢𝑛𝑑𝑠𝑖𝑡−𝑝+ 𝛽6𝑄𝑢𝑎𝑙𝑖𝑡𝑦 ∗ 𝐸𝑆𝐼𝐹𝑢𝑛𝑑𝑠𝑖𝑡−𝑝+ 𝜀𝑖𝑡
9 requires research of its own. For that reason, I will not add it to my analysis. The EMFF represents a far smaller part of the budget and its only relevant for coastal countries. Since my analysis will feature many landlocked countries I will leave this fund outside the scope of my research as well. Then finally the CF has the explicit goal of convergence of member states. However, the current recipients of the cohesion policy are all the new member countries post-2004 plus Greece. None of the EU-15 receive cohesion policy (except Greece).4 To be able to adequately compare the CEE dataset and the full country dataset, which includes the EU-15, I will leave this out of my research as well. For additional information on the European Structural and Investment Funds see Appendix C. The ERDF and ESF data is available through the European Commission’s Database for regional policy. The ESI variable equals the total of ERDF + ESF budgets committed per country per year divided by real GDP per year. Real GDP is taken from the World Bank Development Indicators. Institutional quality will be measured through a number of different proxies: openness to trade, inflation, budget balance and the World Bank’s Worldwide governance indicators. As Sachs et. al (1995) set out, trade can be a proxy for good institutions. However, I will not employ their constructed index as it is not available for the period under consideration, nor would it be useful as there would be almost no variation amongst the CEE countries in degrees of openness. However, their argument for the inclusion of an openness variable still holds. I will, therefore, add the most common proxy for a country’s openness, trade as a percentage of GDP. I do note the limitations of that proxy. The trade variable can be used as a proxy for good institutions through the assumption that the more open a country is, the better its institutions are. However, there are some discrepancies among markets. Some economies lean far more on export than their domestic demand. Take Germany and Austria as an example. In any given year Austria’s trade as a percentage of GDP is far
10 higher than Germany. Which fits Austria’s status as a small landlocked country surrounded by free trade partners and Germany’s status as the EU country with the largest population. Does this mean however that Germany has less good institutions than Austria? I assume not. However, I do believe that, until a certain threshold, trade can be a good proxy for institutional quality. Especially, in the CEE countries whose fast economic development since their accession to the EU trade growth went hand in hand with the adaptation of better policies. For these countries, the argument of trade and openness will still hold and is useful if used as a robustness check amongst other institutional quality indicators. Following Easterly (2005) and Acemoglu et. al (2003) I will add inflation and budget deficit as institutional proxies. That is the annual percentage change in consumer prices taken from the World Bank Development Indicators and the General Government Deficit as a percentage of GDP taken from the OECD. The corruption argument held by EGN can be best measured with the world governance indicator by Kaufmann, Kraay and Zoido-Lobáton (2002), control of corruption, defined as the exercise of public power for private gain. Furthermore, as a robustness check of the general quality of institutions, I will estimate three other WGI indicators (political stability, regulatory quality and government effectiveness) as well. All variables are taken on a yearly per country basis and the ESI funds will be transformed into its natural logarithm for easier interpretation.
I expect there to be large differences in variables across the years. Take, for example, 2007 to 2008. Where the world went from explosive growth to the biggest recession in decades. Therefore I will use time fixed effects by adding year dummies to account for large differences between years.
11 argument goes that countries closer to Brussels receive more funds because the distribution of funds follows a path from center to periphery. However, I do not fully agree with the argument that distance to Brussel is independent of GDP growth. Being closer to Brussel means to some extent being closer to Germany and France. Major markets that drive trade and therefore influence GDP growth. Also, after a first viewing of the data, it can be reasoned that their center-periphery argument does not hold. Finland received ten times less ERDF funds per capita than Estonia in 2012. Their capitals are 82 kilometers apart. An insignificant difference in distance when taking into account that they are about ~1600 kilometers away from Brussels. The Czech Republic in the same year received more than three times the funds per capita as Bulgaria. It is hard to find suitable instruments when your independent variables is GDP growth.
I would, however, propose a new instrument. Bouvet and Dall’Erba (2010) find that the amount of EU-skeptics in the population affect ESI fund allocation. I will employ their proxy variable for skepticism as an instrument. Eurobarometer conducts a yearly survey in which they ask people whether their countries membership to the EU is either: ‘a good thing’, ‘neither good nor bad’ or ‘a bad thing’. The proxy variable will be the percentage of people that answered the latter. Bouvet and Dall’Erba (2010) also find that the orientation of the government, left-wing or rightwing, influences fund allocation. I will not use the political orientation of the government as an instrument because left vs right wing proxy variables are often only used in relation to long-established democracies. The idea is that political division within a country takes time to become fully grounded and can, therefore, offer no benefit for recent democracies like the CEE countries. The 2SLS estimation will take the following form:
(1) 𝑅𝑒𝑔𝐹𝑢𝑛𝑑𝑠̂ 𝑖𝑡 = 𝛽̂ + 𝛽0 ̂𝐸𝑈𝑠𝑐𝑒𝑝𝑡𝑖𝑐𝑖𝑠𝑚 + 𝜀1 𝑖𝑡
(2) 𝑔𝑖𝑡 = 𝛽0+ 𝛽1𝑔𝑖𝑡−1+ 𝛽2𝑠𝑖𝑡 + 𝛽3ℎ𝑢𝑚𝑖𝑡+ 𝛽4(𝑛𝑖𝑡+ 𝛿 + 𝑔𝑎) +
12 I believe my instrument to be strong but to conduct a more robust analysis I will also run a two-step Arellano and Bond (1991) GMM estimation. In the Arellano-Bond two-step system GMM estimation, the lagged values of the endogenous variables are added to the equation as instruments for the endogenous variable. This is done because the endogenous variables may be correlated with the error term. Adding their lagged values as instruments mitigates this bias (Arellano and Bond (1991), Roodman (2006), Mileva (2007)).
I will collect a dataset of all European countries (EU-25) excluding Bulgaria, Romania and Croatia for the time period 2004-2016. I start off with a panel of the 10 CEE countries (EU-10) to estimate the effect of ESI funds on growth with four regressions using three different estimation methods: a country and time fixed effects regressions, a system GMM estimation in which the lagged values of ESI funds are used as instruments, and two 2SLS estimations using EU-skepticism and distance-to-Brussels as instruments respectively.5 It could be argued that distance-to-Brussels is not a good instrument for the 10 accession countries because the distance will be similar for most CEE countries. However, this can only be argued for the EU-10 dataset and table 1 shows that those countries also feature large differences in distance to Brussels. For the full EU-25 data set the differences are of course extensive.
Table 1 Summary of the distance to Brussels variable for the EU-10 dataset from 2004 - 2016
I will subsequently move on to test whether the findings of EGN, who used a panel of 13 EU countries for the period 1960 to 1995, translate to my full 25
5 EU-10 (all countries that acceded to the EU in 2004): Estonia, Latvia, Lithuania, Poland, Czech Republic, Slovakia, Hungary, Slovenia, Cyprus, Malta.
Variable Observations Mean Std. Dev. Min Max
13 country data set for the period 2004-2016.6 It will be interesting to see if these
results are similar, because the global financial crisis and subsequent Euro crisis played a large role in my period. Mohl and Hagen (2010) found that the effect of ESI funds are negative at multiple lags when estimating with a two-step system GMM method. It is necessary to estimate the effects at different lags because it might take time for ESI funds to influence economic growth. Think for example of an infrastructure project. A government wants to use EU funds to connect to of its major cities with a highway. It is probable that such a project is not completed in a year and, therefore, will not benefit growth in the same year. Mohl and Hagen (2010) find that the signs of the ESI funds change at different lags. I will test this counterintuitive finding on my full dataset and the CEE dataset as well. I will then extend on the literature by identifying the effect of ESI funds interaction with institutional quality at different lags for the full EU-25 dataset and the EU-10 dataset. I have chosen a country-level analysis over a region-level analysis because there are no accurate indicators for institutional quality at the regional-level available. The priorities of ERDF and ESF funds are decided on in the multiannual financial frameworks, therefore there might be discrepancies between the data from 2004-2006, 2007-2012 and 2013-2016. For example, the three objectives of the ERDF funds in the 2000-2006 period named in section II are not the same as the two objectives of the 2007-2012 period (Regional Competitiveness and Convergence). However, I believe these differences will be arbitrary because these objectives are ‘thematic’ with an overarching goal of economic growth. Thematic in the sense that they are based on different types of driving economic growth. For that reason, I will pool all data from all three periods in my main analysis.
14 IV. Regression Results
The results of a modified Wald test for group-wise heteroscedasticity in a fixed-effects regression indicate the presence of heteroscedasticity in my panel. Furthermore, the null hypothesis of a Wooldridge test for serial correlation is also rejected. Therefore robust standard errors will be used in all models. Table 1 reports the results of the country and time fixed effects estimation of the augmented Solow-Swan model for EU-10 countries from 2004 to 2016. The savings rate and the variable for population growth, depreciation and technological progress both take their expected signs and are significant. Human capital’s effect on growth is positive and significant in the 2SLS estimation and insignificant in the fixed effects and system GMM estimations. More importantly, the effect of ESI funds is negative and significant. This significant negative effect is robust for the system GMM estimation as well as the 2SLS estimations with skepticism and distance as instruments respectively. There is no large difference between using skepticism and distance. The argument that distance-to-Brussels is less applicable for the EU-10 countries because there is too little variation in distance does not hold. It should have led to a large discrepancy between regressions for the EU-10 and EU-25 datasets. As columns 4 and 5 of table 2 show, no such discrepancy exists. To check the strength of my instrument I use the test based on the F-statistic of the first stage regression as set out by Staiger and Stock (1994). For all three 2SLS regressions the instruments the p-value is below the 5% significance level. To check the exogeneity of my instruments I ran the Sargan (1958) test for over-identifying rejections. The p-values are high and I thus fail to reject the null that the over-identifications are valid.
15 amount between 50 to 85 percent of the total project costs. A higher amount of funds received from the EU means a higher amount of rent, but also a higher amount of national and regional funds required for the project. This corruption hypothesis can be considered quite a stretch and therefore requires testing.
Table 2. Time & Country Fixed Effects, System GMM and 2SLS with skepticism and distance as instruments.
For that reason, I will now move on to test the hypothesis that the positive effects of ESI funds are influenced when they are interacted with the quality of a country’s institutions. In table 3 I regress the effect of ESI funds on growth while including an interaction for different indicators of good governance. I only estimate using system GMM, because I believe it is safe to assume that the institutional quality indicators themselves also suffer from endogeneity and I do not have instruments for all quality indicators. Therefore I will not employ 2SLS for the regressions that include interactions. As mentioned in the methodology section the Arellano and Bond (1991) two-step
(1) (2) (3) (4) (5)
FE SGMM 2SLS
Instrument ESIit-1/GDPit-1 Skepticism Distance Distance
VARIABLES git git git git git
git-1 0.308*** 0.327*** -0.0200 0.455*** 0.259** (0.0878) (0.0675) (0.145) (0.0589) (0.118) sit 0.373*** 0.237*** 0.809*** 0.255** 0.495** (0.122) (0.0489) (0.206) (0.109) (0.224) nit + δ + ga -1.214*** -2.329** -4.534*** -1.362*** -2.025*** (0.366) (0.969) (1.320) (0.475) (0.732) humit 0.0483 -0.0294 0.331** 0.100* 0.0882 (0.0515) (0.0243) (0.136) (0.0594) (0.0868) Log ESIit/GDPit -1.124** -0.190*** -5.340** -1.112*** -1.029*** (0.495) (0.055) (2.190) (0.313) (0.331) Sargan Test 0.898 0.572 0.532 F – Test 0.0155 0.0242 0.0359 Observations 130 130 130 130 325 R-squared 0.732 # Countries 10 10 10 10 25
16 system GMM estimator is used widely in the literature as the next best solution when instruments for a two-stage-least-square estimation are unavailable. The results of the estimation show that all variables roughly keep their sign and significance. ESI funds still exert a negative influence on economic growth. However there are significant interaction effects with four institutional quality indicators, trade, inflation, budget and political stability. The interaction with inflation is supposed to be negative. Pursuing higher interest rates is considered good governance in this context. Although these interactions have a positive sign and therefore indicate a mitigating response on ESI funds’ negative effects, they are not of large enough size to truly contribute. A country would need extraordinary openness or negative inflation. Political stability would also need to be on the high end if it were to outweigh the negative effects of ESI funds. The funds will therefore most likely not display positive effects in the first year.
Table 3. ESI funds conditional on institutions.
(1) (2) (3) (4) (5) (6) (7)
Interaction Trade Inflation Budget CoC GE PS VaA
Dependent g g g g g g g git-1 0.29*** 0.354*** 0.22*** 0.29*** 0.29*** 0.289*** 0.30*** (0.104) (0.117) (0.0766) (0.102) (0.104) (0.0979) (0.093) sit 0.115 0.192** 0.121 0.227** 0.244** 0.204** 0.23** (0.0892) (0.0863) (0.1000) (0.109) (0.0977) (0.0933) (0.109) nit + δ + ga -2.27** -2.281*** -2.30*** -2.33** -2.31*** -2.89*** -2.31** (0.898) (0.859) (0.840) (0.935) (0.871) (0.945) (0.947) humit 0.0221 -0.0282 -0.0109 -0.0344 -0.0279 -0.0374 -0.0185 (0.0200) (0.0254) (0.0288) (0.033) (0.0317) (0.0246) (0.029) Log ESIit -5.73** -1.555 -0.890 -2.20** -1.273 -5.555*** -2.160 (2.683) (0.954) (0.816) (0.878) (1.920) (1.674) (1.426) Interaction 0.238* -0.213* 3.95*** -0.434 -1.194 4.406*** -0.0869 (0.134) (0.113) (1.08) (1.014) (2.207) (1.338) (1.432) Observations 130 130 130 130 130 130 130 # Countries 10 10 10 10 10 10 10
17 Table 3 does not offer a robust conclusion to the corruption hypothesis. Only four out of seven interaction effects are significant. Even in the case that they were enough significant interaction effects, their coefficients are of insufficient magnitude to compensate the negative effect of ESI funds. Positive effects may still arise over time. Therefore, I will now test the effect of the funds while lagging up to four years. In this case I will use my instrumental variable again, because I will not be interacting with institutional variables and I want to test whether the findings are robust for both the 2SLS and system GMM methods. Table 4 shows the results for the EU-25 dataset, while table 5 shows the results for the EU-10 dataset. In the EU-10 countries the system GMM method offers significant positive effects of ESI funds for one-, two- and three-year lags. The negative effect without a lag and the positive effect of the one-year lag are robust for the two-stage-least-squares estimation. In the EU-25 dataset there are fewer significant effects for those two lags, but the majority of the coefficient take similar sign and magnitude.
18 Table 4. EU-25 dataset regressions lagged up to four years
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
SGMM 2SLS SGMM 2SLS SGMM 2SLS SGMM 2SLS SGMM 2SLS
Variables git git git git git git git git git git
git-1 0.327*** -0.159 0.333*** -0.0135 0.391*** 0.136 0.322*** 0.133 0.133** 0.00983 (0.0675) (0.177) (0.0588) (0.166) (0.0732) (0.0980) (0.0603) (0.138) (0.0592) (0.0977) sit 0.237*** 0.842*** 0.181*** 0.994*** 0.175*** 0.835*** 0.199*** 0.752*** 0.201*** 0.660* (0.0489) (0.167) (0.0583) (0.296) (0.0468) (0.192) (0.0552) (0.252) (0.0694) (0.375) nit + δ + ga -2.329** -1.700*** -1.447** -2.233*** -1.489** -2.037*** -1.096* -2.317*** -0.590 -2.399*** (0.969) (0.611) (0.678) (0.803) (0.680) (0.613) (0.596) (0.592) (0.631) (0.559) humit -0.0294 -0.0823 -0.0130 -0.139 -0.0130 -0.0532 -0.0191 -0.0175 -0.0144 0.0337 (0.0243) (0.112) (0.0199) (0.145) (0.0154) (0.144) (0.0177) (0.150) (0.0202) (0.524) Log ESIit/GDPit -0.790 -7.420** -3.35*** -3.856 -4.116*** -3.450* -3.914*** -2.237* -4.695*** -1.432 (0.655) (3.667) (0.669) (4.232) (0.477) (2.080) (0.485) (1.132) (0.518) (4.762) Log ESIit-1/GDPit-1 3.359*** 0.0534 6.103*** 4.348 5.979*** 4.772** 6.335*** 3.245 (0.685) (2.877) (0.742) (5.742) (0.643) (2.329) (0.673) (4.322) Log ESIit-2/GDPit-2 2.135*** -3.180 1.331** -7.559 -0.958 0.141 (0.399) (5.633) (0.522) (8.811) (1.036) (2.264) Log ESIit-3/GDPit-3 0.137 6.066 1.662 -4.430 (0.842) (8.709) (1.069) (18.85) Log ESIit-4/GDPit-4 -1.640 6.049 (1.752) (25.10) Observations 325 325 300 300 275 275 250 250 225 225 # Countries 25 25 25 25 25 25 25 25 25 25
19 Table 5. EU-10 dataset regressions lagged up to four years
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
SGMM 2SLS SGMM 2SLS SGMM 2SLS SGMM 2SLS SGMM 2SLS
Variables git git git git git git git git git git
git-1 0.451*** 0.178* 0.475*** 0.252** 0.453*** 0.298*** 0.384*** 0.235** 0.277*** 0.139 (0.0375) (0.0957) (0.0453) (0.100) (0.0514) (0.102) (0.0757) (0.103) (0.0829) (0.0945) sit 0.178*** 0.818*** 0.0996 0.598*** 0.0860 0.796*** 0.0771 0.995*** 0.0466 0.642*** (0.0588) (0.158) (0.0618) (0.187) (0.0671) (0.205) (0.0648) (0.225) (0.0463) (0.230) nit + δ + ga -1.45*** -4.760*** -0.531 -4.846*** -0.305 -4.719*** 0.0918 -4.326*** 0.679 -3.022** (0.528) (1.300) (0.511) (1.279) (0.627) (1.319) (0.698) (1.370) (0.499) (1.316) humit -0.0443* 0.248 -0.0212 0.488** -0.0216 0.267 -0.0197 0.0160 -0.0136 0.0113 (0.0230) (0.189) (0.0254) (0.218) (0.0263) (0.259) (0.0275) (0.288) (0.0265) (0.298) Log ESIit/GDPit -0.539 -2.184*** -1.85*** -3.776*** -1.818*** -3.369*** -1.733*** -2.209** -1.812*** -2.632** (0.605) (0.709) (0.634) (1.021) (0.618) (1.068) (0.650) (1.102) (0.672) (1.020) Log ESIit-1/GDPit-1 2.704*** 2.483** 2.691*** 4.691*** 2.660*** 4.503*** 2.850*** 5.695*** (0.719) (1.165) (0.638) (1.507) (0.521) (1.480) (0.670) (1.371) Log ESIit-2/GDPit-2 0.196 2.527** 0.714** -2.058 1.027*** -0.727 (0.590) (1.173) (0.316) (1.460) (0.335) (1.334) Log ESIit-3/GDPit-3 -0.0849 -0.436 0.998* 1.247 (0.654) (1.181) (0.551) (1.302) Log ESIit-4/GDPit-4 -0.893 -1.378 (0.676) (1.188) Observations 130 130 120 120 110 110 100 100 90 90 # Countries 10 10 10 10 10 10 10 10 10 10
20 The interaction term with inflation is expected to be negative because low inflation is used as a proxy for good governance. The fact that budget interacts positively with ESI funds is not surprising, Bouvet and Dall’erba (2010) showed that wealthy countries with large budgets have an easier time co-financing and are therefore more likely to receive larger amounts of ESI funds.
Table 6. Coefficients for ESI and its interaction with Institutional Quality for Arellano Bond System GMM Estimation for the 10 CEE countries over the period 2004 – 2016.
(1) (2) (3) (4) (5) (6) (7) (8)
Interaction Trade Inflation Budget CoC GE PS RQ VaA Dependent Variable git git git git git git git git ESI T-1 -2.460 0.309 0.200 -0.269 -0.255 -2.040** 0.539 -1.220 (1.710) (0.456) (0.456) (0.606) (1.06) (0.995) (0.90) (0.876) Interaction 0.015 -0.22*** 0.209** 0.198 0.099 2.470*** -0.663 1.120* (0.010) (0.087) (0.083) (0.655) (1.292) (0.947) (0.66) (0.596) AR(1) 0.005 0.003 0.005 0.005 0.007 0.005 0.007 0.007 JS Test 0.333 0.008 0.039 0.901 0.933 0.033 0.567 0.169 Observation 120 120 120 120 120 120 120 120 Countries 10 10 10 10 10 10 10 10 ESI T-2 -2.240 0.197 0.281 0.013 -0.030 -1.340 0.674 -0.571 (1.740) (0.430) (0.449) (0.626) (1.04) (0.928) (0.874) (1.060) Interaction 0.015 -0.172 0.166* -0.057 0.018 1.830** -0.692 0.586 (0.011) (0.122) (0.096) (0.506) (1.12) (0.884) (0.57) (0.700) AR(1) 0.005 0.004 0.008 0.005 0.006 0.005 0.007 0.007 JS Test 0.393 0.194 0.205 0.993 0.999 0.107 0.453 0.616 Observation 110 110 110 110 110 110 110 110 Countries 10 10 10 10 10 10 10 10 ESI T-3 0.051 0.817** 0.876** 1.230** 1.210 -0.874 1.830* 0.514 (1.310) (0.353) (0.389) (0.592) (1.16) (1.110) (1.08) (1.260) Interaction 0.005 -0.139** 0.056 -0.752 -0.420 2.290** -0.999 0.272 (0.008) (0.065) (0.091) (0.637) (1.26) (1.030) (0.85) (1.180) AR(1) 0.009 0.006 0.010 0.009 0.013 0.007 0.014 0.014 JS Test 0.086 0.000 0.065 0.106 0.046 0.000 0.168 0.136 Observation 100 100 100 100 100 100 100 100 Countries 10 10 10 10 10 10 10 10 ESI T-4 1.000 2.100*** 1.81*** 2.63*** 2.830* 0.322 3.04** 1.830 (1.570) (0.525) (0.444) (0.747) (1.541) (1.510) (1.256) (1.490) Interaction 0.006 0.208 -0.198 -1.170 -0.926 2.080 -1.220 0.105 (0.010) (0.153) (0.120) (0.792) (1.591) (1.460) (1.012) (1.571) AR(1) 0.006 0.007 0.008 0.006 0.007 0.006 0.005 0.010 JS Test 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.000 Observation 90 90 90 90 90 90 90 90 Countries 10 10 10 10 10 10 10 10
21 Table 7. Coefficients for ESI and its interaction with Institutional Quality for Arellano Bond System
GMM Estimation for the full dataset of 25 countries over the period 2004 – 2016.
(1) (2) (3) (4) (5) (6) (7) (8)
Interaction Trade Inflation Budget CoC GE PS RQ VaA Dependent variable git git git git git git git git Log ESIit -3.39*** -0.301 0.590 -0.654 -0.897 -2.87*** -0.809 -1.320 (0.970) (0.633) (0.839) (0.78) (1.11) (0.917) (0.85) (1.030) Interaction 0.019*** -0.173* 0.37*** 0.012 0.377 2.960*** 0.101 0.925 (0.004) (0.102) (0.105) (0.90) (1.28) (0.883) (0.97) (1.170) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.102 0.001 0.513 0.411 0.003 0.267 0.320 Observation 325 325 325 325 325 325 325 325 Countries 25 25 25 25 25 25 25 25 LogESI T-1 -2.21*** 0.691** 1.030* 0.059 -0.039 -0.210 -0.530 -0.492 (0.834) (0.328) (0.548) (0.60) (0.95) (0.578) (0.71) (0.820) Interaction 0.016*** -0.20*** 0.27*** 0.470 0.592 0.947 0.796 1.060 (0.004) (0.071) (0.079) (0.70) (1.13) (0.642) (0.76) (0.912) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.005 0.002 0.604 0.291 0.121 0.568 0.283 Observation 300 300 300 300 300 300 300 300 Countries 25 25 25 25 25 25 25 25 LogESI T-2 -2.410** 0.612** 1.040** 0.206 0.762 0.113 0.013 0.154 (1.010) (0.297) (0.465) (0.71) (0.95) (0.730) (0.72) (1.210) Interaction 0.017*** -0.158* 0.26*** 0.394 0.093 0.696 0.338 0.580 (0.005) (0.085) (0.073) (0.66) (1.01) (0.765) (0.63) (1.180) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.001 0.000 0.422 0.007 0.071 0.674 0.143 Observation 275 275 275 275 275 275 275 275 Countries 25 25 25 25 25 25 25 25 LogESI T-3 -2.040* 1.100*** 0.998** 0.989 1.410 0.523 0.928 1.100 (1.180) (0.236) (0.408) (0.83) (0.96) (0.682) (0.83) (1.340) Interaction 0.018*** -0.189 0.20*** -0.374 -0.178 0.781 -0.123 -0.161 (0.006) (0.116) (0.073) (0.72) (0.99) (0.605) (0.70) (1.370) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.000 0.001 0.220 0.000 0.000 0.155 0.035 Observation 250 250 250 250 250 250 250 250 Countries 25 25 25 25 25 25 25 25 LogESI T-4 -2.490 1.160*** 0.945 2.08** 3.31** 1.420 1.320 2.480 (1.730) (0.336) (0.602) (0.98) (1.63) (1.010) (1.03) (1.850) Interaction 0.025** -0.351* 0.142 -1.380 -1.950 -0.051 -0.175 -0.988 (0.010) (0.197) (0.105) (0.92) (1.48) (1.070) (0.76) (1.800) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.000 0.197 0.069 0.001 0.654 0.145 0.101 Observation 225 225 225 225 225 225 225 225 Countries 25 25 25 25 25 25 25 25
22 The interaction effect is only robust for two World Bank Governance Indicators. For example, a country with political stability will be better equipped to guarantee the completion of long-term investments.
When adding an interaction effect to a regression the interpretation from the included variables changes. The effect of ESI funds on growth will now also depend on the institutional quality. This is why ESI’s coefficient can still be negative and why in table 6 and 7 the coefficients with trade are relatively large. The mean of the observations for trade as a percentage of GDP is higher than, for example, the mean of the observations for budget deficit as a percentage of GDP. The sign of the interaction is what is most relevant with interaction effects and some rudimentary conclusions may only be drawn from the magnitude of the coefficient after plugging in representative values. To understand the interpretation of the coefficients in the interaction regressions let me illustrate with an example. Take the regression that includes political stability (table 5, column 6, the first row of coefficients). Assume a fictional EU member state, Europhilia, which receives an average amount of ESI funds of 0.5 percent of its GDP and has an above average level of political stability of 1 (see Appendix B for summary statistics). The effect that the ESI funds, conditional on institutions, will have on growth is: 0.5 ∗ −2.87/100 + 0.5 ∗ 1 ∗ 2.960/100 = 0.0045.7 Say that Europhilia effectively lobbies for more funds in the next
multiannual financial framework and the amount of ESI funds it receives increases to 0.7 percent of its GDP. The effect on economic growth will now be 0.7 ∗ −2.87/100 + 1 ∗ 0.7 ∗ 2.960/100 = 0.0063. The difference in effect between both scenarios is 0.063 − 0.045 = 0.00585. So as a rough interpretation, for a country which has a political stability index of 1 and the amount of ESI funds it receives increases from 0.5 percent of its GDP to 0.7 percent than its economic growth rate will increase with 0.0585 percentage points as a result from that. Looking at the effect equation it can clearly be seen
23 that if Europhilia would have had a lower level of political stability, the positive interaction would not outweigh the negative effect. Thus receiving ESI funds would lead to a decrease in economic growth.
24 Figure 1. Predicted values of growth for deficits ranging from -3 to 3 percent and the amount of ESI funds received from 1 to 3 percent of GDP. Based on the regressions shown in table 6, ESI T-2.
V. Discussion
25 Solow model is that without the addition of human capital the magnitude of the effects of population growth and savings were amplified, although their signs are correctly measured. So even though I added a measure of human capital to my regressions it is only significant in the 2SLS and time and country fixed effects regressions. It is, therefore, possible that in the other regressions the signs of the effect of savings and population growth, depreciation and technological progress are upward biased. This bias does not affect my main variable of interest, ESI funds, because I assume that human capital and the amount of funds a country receives are uncorrelated.
The focus of my paper was to estimate what effect ESI funds have on growth in the CEE countries and the effect ESI funds have in interaction with institutional quality. In contrast to Mohl and Hagen (2010) who only found significant positive effects for the objective 1 part of the ERDF funds. When summing all objectives they found changing signs at different lags. They argue that this effect is the result objectives 2 and 3 not having clear guidelines. My results are more intuitive, possibly because the guidelines of the funds are more clear cut in my time frame (see Appendix C for a summary of the funds). In my EU-10 country dataset, there is a negative effect of ESI funds when it’s not lagged. However, starting at three-year lags, the funds start to show positive effects on economic growth. The result that ESI funds are negative without a lag and positive with a one-year lag are robust for the two-stage-least-squares estimation.
27 allow me to draw a more robust conclusion. EGN found a significant interaction effect for most of their institutional quality variables, this result does not translate to my timeframe and country sets. A few of my interaction coefficients have significant results that point in the direction of the results of EGN, but because there is no large majority of significant effects for more than half of the institutional quality indicators for the regressions with lags, I conclude that the effect of ESI funds are not influenced by institutional quality, neither in the EU-25 nor the EU-10 dataset.
VI. Conclusion
This paper set out to analyze the effects of ESI funds on economic growth for the most recent time period (2006-2016) and the 2004 EU accession countries. I aimed to analyze what the effect was at different lags and what role institutional quality plays in their effectiveness. In line with Mohl and Hagen (2010) I found significant positive effects of the funds after lagging one year, which is robust for both the System GMM estimation and the 2SLS estimation with my own proposed instrument of EU-skepticism. The system GMM estimation also offers positive significant effects at two and three-year lags in my EU-10 dataset. Extending on the literature, I estimated ESI funds also interacting with institutional quality at different time lags. When adding interaction effects with institutional quality indicators in these lagged regressions there are only a few significant interactions, but the effect of ESI remains positive. Although the results point to some significant interaction effects they are not robust for more than half of the indicators. Whether institutional quality influences the effect of ESI funds on growth is, therefore, ambiguous.
28 current members. My findings point out that the role of institutional quality in the effectiveness of the ESI funds in the CEE countries is not significant.
References
Acemoglu, Daron, Simon Johnson, James Robinson, and Yunyong Thaicharoen. "Institutional causes, macroeconomic symptoms: volatility, crises and growth." Journal of monetary economics 50, no. 1 (2003): 49-123.
Arellano, Manuel, and Stephen Bond. "Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations." The review of economic studies 58, no. 2 (1991): 277-297.
Becker, Sascha O., Peter H. Egger, and Maximilian Von Ehrlich. "Going NUTS: The effect of EU Structural Funds on regional performance." Journal of Public Economics 94, no. 9-10 (2010): 578-590.
Bouvet, Florence, and Sandy Dall'Erba. "European regional structural funds: How large is the influence of politics on the allocation process?." JCMS: Journal of Common Market Studies 48, no. 3 (2010): 501-528.
Bräutigam, Deborah. Aid dependence and governance. Vol. 1. Stockholm: Almqvist & Wiksell International, 2000.
Burnside, Craig, and David Dollar. "Aid, policies, and growth." American economic review 90, no. 4 (2000): 847-868.
29 De Haan, Jakob, and Jan-Egbert Sturm. "Political and institutional determinants of fiscal policy in the European Community." Public Choice 80, no. 1-2 (1994): 157-172.
Easterly, William. "National policies and economic growth: a reappraisal." Handbook of economic growth 1 (2005): 1015-1059.
Ederveen, Sjef, Henri LF De Groot, and Richard Nahuis. "Fertile soil for structural funds? A panel data analysis of the conditional effectiveness of European cohesion policy." Kyklos 59, no. 1 (2006): 17-42.
Inglehart, Ronald, and Pippa Norris. "Trump, Brexit, and the rise of populism: Economic have-nots and cultural backlash." (2016).
Kaufmann, D. "A. Kray y P. Zoido-Lobaton.“Governance Matters II”." Policy Research Working Paper 2772 (2002).
Mankiw, Gregory N., David Romer and David N. Weil. "A contribution to the empirics of economic growth." The quarterly journal of economics 107 (1992): 407-437.
Mileva, Elitza. "Using Arellano-Bond dynamic panel GMM estimators in Stata." Economics Department, Fordham University (2007): 1-10.
Mohl, Philipp, and Tobias Hagen. "Do EU structural funds promote regional growth? New evidence from various panel data approaches." Regional Science and Urban Economics 40, no. 5 (2010): 353-365.
30 Roodman, David. "How to do xtabond2: an introduction to ‘difference’and ‘system." In GMM in STATA’, Center for Global Development Working Paper No. 103. 2006.
Rutte, Mark. “Underpromise and Overdeliver: fulfilling the promise of Europe.” Speech, Bertelsmann Stiftung, Berlin, March 02, 2018.
Sachs, Jeffrey D., Andrew Warner, Anders Åslund, and Stanley Fischer. "Economic reform and the process of global integration." Brookings papers on economic activity 1995, no. 1 (1995): 1-118.
Sargan, John D. "The estimation of economic relationships using instrumental variables." Econometrica: Journal of the Econometric Society (1958): 393-415. Schaffer, Mark. "xtivreg2: Stata module to perform extended IV/2SLS, GMM and AC/HAC, LIML and k-class regression for panel data models." (2015). Solow, Robert M. "A contribution to the theory of economic growth." The quarterly journal of economics 70, no. 1 (1956): 65-94.
Staiger, Douglas O., and James H. Stock. "Instrumental variables regression with weak instruments." (1994).
Swan, Trevor W. "Economic growth and capital accumulation." Economic record 32, no. 2 (1956): 334-361.
Appendix A.
Regressions without Hungary and Poland.
(1) (2) (3) (4)
FE SGMM 2SLS
Instrument ESIit-1/GDPit-1 Skepticism Distance
VARIABLES git git git git
31 (0.128) (0.0524) (0.260) (0.118) nit + δ + ga -1.263*** -2.361** -4.652*** -1.443*** (0.374) (1.045) (1.766) (0.488) humit 0.0300 -0.0409 0.129 0.0869 (0.0508) (0.0250) (0.165) (0.0594) ESIit/GDPit -0.423 -2.13*** -3.97* -2.11*** (0.511) (0.620) (2.220) (0.367) Observations 96 96 96 96 R-squared 0.777 # Countries 8 8 8 8
Robust standard errors in parentheses *** p<0.01, ** p<0.05, *
32 Appendix B. Summary Statistics
Variable Mean Std. Dev. Min Max Observations
33 within 0.163424 0.226842 1.329442 T = 13 RQ overall 1.286912 0.371809 0.148387 1.976167 N = 325 between 0.356169 0.677546 1.789596 n = 25 within 0.126928 0.757753 1.629296 T = 13 RoL overall 1.262802 0.509358 0.107359 2.100273 N = 325 between 0.505998 0.446227 1.977834 n = 25 within 0.113793 0.754505 1.560449 T = 13 VaA overall 1.20337 0.266244 0.401232 1.800992 N = 325 between 0.255229 0.797929 1.594328 n = 25 within 0.09039 0.736342 1.5191 T = 13 Appendix C.
A Summary of the ESIF Funds.
There are five different European Stability and Investment Funds: ERDF, ESF, CF, EARDF and EMFF. The purpose of all funds is investment towards the creation of jobs and a sustainable economy and environment. However different funds take different approaches based on thematic objectives. Thematic objectives and funding are decided on in the multiannual financial frameworks. This means that although the overall goals of the funds stay the same, the way of reaching these objectives may change. The European Regional Development Fund (ERDF) focusses on reducing economic disparities between regions. In the different frameworks this was done through:
2000 – 2006:
1. Supporting development in less prosperous regions 2. Revitalizing areas facing structural difficulties. 2007 – 2013:
34 Innovation and research, the digital agenda, support for small and medium-sized enterprises and the low-carbon economies.
There are guidelines for funds allocation in the 2006-2013 framework. A percentage of funds must focus on at least two priorities. In the developed regions, transition regions and less developed regions this is 80 percent, 60 percent and 50 percent respectively. While at least 20 percent, 15 percent and 12 percent respectively must specifically be allocated towards the low-carbon economy.
The European Social Investment Fund “invests in people”. The overall goal of it is to promote employment and support labor mobility. In the 2014-2020 period, this is done through: supporting labor mobility, promoting social inclusion and combating poverty, investing in education, skills and lifelong learning, enhancing institutional capacity and an efficient public administration. The Cohesion Fund is there only for member states with a GNI which is less than 90 percent of the EU average Investment is done into environment and transport networks. The European Agricultural Fund for Rural Development is accessible for all member states and focusses on agricultural competitiveness. The European Maritime and Fisheries Fund focusses on job creation in the maritime industry.
35 Appendix D.
Regressions for the Western EU countries.
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git ESIit/GDPit 0.488 -0.761 2.290 -1.05* -1.240 -1.55** -3.55*** 4.48*** (1.96) (0.671) (1.670) (0.61) (0.78) (0.62) (0.90) (1.56) Interaction -0.017 0.082 0.38*** 0.736 0.562 1.670* 3.63*** 3.90*** (0.02) (0.185) -0.146) (0.65) (1.01) (0.91) (1.33) (1.45) AR(1) 0.005 0.005 0.008 0.006 0.005 0.006 0.005 0.006 JS Test 0.355 0.525 0.001 0.121 0.197 0.008 0.001 0.015 Observations 195 195 195 195 195 195 195 195 Number of cn 15 15 15 15 15 15 15 15
ESIit-1/GDPit-1 -1.220 -0.208 2.130* -0.745 -1.020 -0.868 -2.11** -1.220
(2.16) (0.539) (1.100) (0.66) (0.86) (0.59) (0.848) (2.16) Interaction 0.013 -0.063 0.33*** 0.912* 0.776 1.200* 2.51*** 0.013 (0.03) (0.171) (0.107) (0.55) (0.79) (0.70) (0.86) (0.03) AR(1) 0.006 0.005 0.008 0.005 0.006 0.005 0.005 0.005 JS Test 0.819 0.864 0.002 0.200 0.489 0.133 0.012 0.116 Observations 180 180 180 180 180 180 180 180 Number of cn 15 15 15 15 15 15 15 15
36 ESIit-2/GDPit-2 -2.730 -0.066 1.730** -0.727 -0.704 -0.666 -1.530* -2.730
(2.39) (0.603) (0.742) (0.70) (0.84) (0.66) (0.807) (2.39) Interaction 0.035 -0.128 0.29*** 0.79** 0.563 0.918* 1.90*** 0.035 (0.02) (0.191) (0.072) (0.40) (0.46) (0.51) (0.729) (0.03) AR(1) 0.008 0.005 0.010 0.005 0.005 0.005 0.006 0.005 JS Test 0.469 0.789 0.000 0.127 0.475 0.205 0.034 0.316 Observations 165 165 165 165 165 165 165 165 Number of cn 15 15 15 15 15 15 15 15
ESIit-3/GDPit-3 -3.460 -0.009 1.010 -1.080 -1.170 -1.040 -1.470* -3.460
(3.47) (0.596) (0.709) (0.76) (0.94) (0.73) (0.830) (3.47) Interaction 0.042 -0.298* 0.29*** 0.88** 0.855 0.900 1.120 0.042 (0.05) (0.176) (0.067) (0.44) (0.58) (0.61) (0.994) (0.05) AR(1) 0.010 0.006 0.012 0.008 0.006 0.006 0.007 0.008 JS Test 0.582 0.178 0.000 0.139 0.328 0.285 0.205 0.252 Observations 150 150 150 150 150 150 150 150 Number of cn 15 15 15 15 15 15 15 15
ESIit-4/GDPit-4 -2.550 -0.790 0.320 -1.920** -1.990* -1.860** -1.550 -2.550
(5.470) (0.775) (0.705) (0.876) (1.120) (0.807) (1.010) (5.470) Interaction 0.017 -0.46*** 0.319*** 0.854 0.820 0.938 -0.101 0.017 (0.078) (0.153) (0.062) (0.780) (1.160) (1.010) (1.480) (0.078) AR(1) 0.010 0.005 0.014 0.009 0.006 0.006 0.007 0.008 Joint Significance 0.289 0.010 0.000 0.086 0.165 0.069 0.149 0.190 Observations 135 135 135 135 135 135 135 135 Number of cn 15 15 15 15 15 15 15 15
37 Regressions for the Western Countries a ESI
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.253** 0.301*** 0.261** 0.297*** 0.291*** 0.276*** 0.293*** 0.296*** (0.0984) (0.106) (0.121) (0.106) (0.0982) (0.0943) (0.0968) (0.0970) sit 0.190*** 0.0986*** 0.0983*** 0.0952*** 0.0908*** 0.133*** 0.0997*** 0.0914*** (0.0447) (0.0284) (0.0372) (0.0331) (0.0301) (0.0353) (0.0367) (0.0287) nit + δ + ga -1.008*** -0.995*** -1.098*** -1.118*** -1.114*** -0.949*** -1.111*** -0.968*** (0.304) (0.256) (0.375) (0.345) (0.364) (0.343) (0.346) (0.284) humit -0.0206 -0.00933 -0.00880 -0.00652 -0.0142 0.000210 -0.0119 -0.00823 (0.0142) (0.0115) (0.0125) (0.0114) (0.0124) (0.0103) (0.00915) (0.0126) Log ESIit/GDPit 0.488 -0.761 2.290 -1.05* -1.240 -1.55** -3.55*** 4.48*** (1.96) (0.671) (1.670) (0.61) (0.78) (0.62) (0.90) (1.56) Interaction -0.017 0.082 0.38*** 0.736 0.562 1.670* 3.63*** 3.90*** (0.02) (0.185) -0.146) (0.65) (1.01) (0.91) (1.33) (1.45) AR(1) 0.005 0.005 0.008 0.006 0.005 0.006 0.005 0.006 JS Test 0.355 0.525 0.001 0.121 0.197 0.008 0.001 0.015 Observations 195 195 195 195 195 195 195 195 # Countries 15 15 15 15 15 15 15 15
38 Regressions for the Western Countries b ESI T-1
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.242** 0.281** 0.244** 0.290*** 0.282*** 0.254*** 0.285*** 0.281*** (0.0980) (0.109) (0.124) (0.101) (0.0991) (0.0982) (0.0958) (0.0958) sit 0.189*** 0.103*** 0.100** 0.104*** 0.100*** 0.161*** 0.107** 0.110*** (0.0633) (0.0320) (0.0394) (0.0385) (0.0376) (0.0440) (0.0457) (0.0364) nit + δ + ga -1.027*** -1.021*** -1.141*** -1.148*** -1.042*** -0.867*** -1.138*** -0.983*** (0.345) (0.242) (0.378) (0.336) (0.364) (0.316) (0.354) (0.296) humit -0.00616 -0.000633 -0.000345 0.00258 -0.000631 0.0191 -0.00298 0.00439 (0.0132) (0.0116) (0.0137) (0.0149) (0.0144) (0.0122) (0.0132) (0.0174) Log ESIit-1/GDPit-1 -1.220 -0.208 2.130* -0.745 -1.020 -0.868 -2.11** -1.220 (2.16) (0.539) (1.100) (0.66) (0.86) (0.59) (0.848) (2.16) Interaction 0.013 -0.063 0.33*** 0.912* 0.776 1.200* 2.51*** 0.013 (0.03) (0.171) (0.107) (0.55) (0.79) (0.70) (0.86) (0.03) AR(1) 0.006 0.005 0.008 0.005 0.006 0.005 0.005 0.005 JS Test 0.819 0.864 0.002 0.200 0.489 0.133 0.012 0.116 Observations 180 180 180 180 180 180 180 180 # Countries 15 15 15 15 15 15 15 15
39 Regressions for the Western Countries c ESI T-2
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.216** 0.260** 0.228* 0.271*** 0.258** 0.219** 0.266*** 0.256** (0.104) (0.114) (0.130) (0.105) (0.104) (0.101) (0.102) (0.110) sit 0.199*** 0.111*** 0.104** 0.118** 0.112*** 0.181*** 0.108** 0.116*** (0.0696) (0.0381) (0.0426) (0.0463) (0.0412) (0.0460) (0.0490) (0.0394) nit + δ + ga -1.064*** -1.098*** -1.261*** -1.335*** -1.141*** -0.818** -1.139*** -1.082*** (0.372) (0.288) (0.483) (0.411) (0.350) (0.347) (0.336) (0.375) humit -0.00283 -0.000547 -0.00336 0.00531 0.00357 0.0265* 0.000733 0.00768 (0.0136) (0.0143) (0.0141) (0.0172) (0.0170) (0.0143) (0.0161) (0.0199) Log ESIit-2/GDPit-2 -2.730 -0.066 1.730** -0.727 -0.704 -0.666 -1.530* -2.730 (2.39) (0.603) (0.742) (0.70) (0.84) (0.66) (0.807) (2.39) Interaction 0.035 -0.128 0.29*** 0.79** 0.563 0.918* 1.90*** 0.035 (0.02) (0.191) (0.072) (0.40) (0.46) (0.51) (0.729) (0.03) AR(1) 0.008 0.005 0.010 0.005 0.005 0.005 0.006 0.005 JS Test 0.469 0.789 0.000 0.127 0.475 0.205 0.034 0.316 Observations 165 165 165 165 165 165 165 165 # Countries 15 15 15 15 15 15 15 15
40 Regressions for the Western Countries d ESI T-3
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.204** 0.218** 0.209* 0.220** 0.222** 0.184* 0.236** 0.213** (0.0995) (0.110) (0.125) (0.0996) (0.0970) (0.0967) (0.102) (0.104) sit 0.176** 0.116*** 0.107** 0.122** 0.118** 0.181*** 0.0965** 0.118** (0.0725) (0.0411) (0.0421) (0.0567) (0.0459) (0.0469) (0.0463) (0.0468) nit + δ + ga -1.225*** -1.325*** -1.417*** -1.666*** -1.380*** -1.067*** -1.472*** -1.453*** (0.430) (0.360) (0.487) (0.599) (0.386) (0.410) (0.361) (0.459) humit -0.0143 -0.00235 -0.0123 -0.00718 -0.00704 0.0143 -0.0172 0.00316 (0.0136) (0.0150) (0.0144) (0.0186) (0.0140) (0.0128) (0.0187) (0.0180) Log ESIit-3/GDPit-3 -3.460 -0.009 1.010 -1.080 -1.170 -1.040 -1.470* -3.460 (3.47) (0.596) (0.709) (0.76) (0.94) (0.73) (0.830) (3.47) Interaction 0.042 -0.298* 0.29*** 0.88** 0.855 0.900 1.120 0.042 (0.05) (0.176) (0.067) (0.44) (0.58) (0.61) (0.994) (0.05) AR(1) 0.010 0.006 0.012 0.008 0.006 0.006 0.007 0.008 JS Test 0.582 0.178 0.000 0.139 0.328 0.285 0.205 0.252 Observations 150 150 150 150 150 150 150 150 # Countries 15 15 15 15 15 15 15 15
41 Regressions for the Western Countries e ESI T-4
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.195* 0.212* 0.215* 0.223** 0.220** 0.204* 0.222** 0.220** (0.104) (0.114) (0.130) (0.109) (0.108) (0.107) (0.109) (0.108) sit 0.167*** 0.106*** 0.104*** 0.104** 0.102** 0.157*** 0.100** 0.113*** (0.0597) (0.0365) (0.0375) (0.0433) (0.0412) (0.0382) (0.0427) (0.0419) nit + δ + ga -1.159*** -1.278*** -1.261*** -1.318*** -1.262*** -1.258*** -1.384*** -1.282*** (0.382) (0.355) (0.370) (0.333) (0.348) (0.333) (0.316) (0.353) humit -0.0140 -0.00643 -0.00640 -0.00709 -0.00832 0.0142 -0.0115 -0.00461 (0.0116) (0.0125) (0.0104) (0.0136) (0.0151) (0.0122) (0.0152) (0.0123) Log ESIit-4/GDPit-4 -2.550 -0.790 0.320 -1.920** -1.990* -1.860** -1.550 -2.550 (5.470) (0.775) (0.705) (0.876) (1.120) (0.807) (1.010) (5.470) Interaction 0.017 -0.46*** 0.319*** 0.854 0.820 0.938 -0.101 0.017 (0.078) (0.153) (0.062) (0.780) (1.160) (1.010) (1.480) (0.078) AR(1) 0.010 0.005 0.014 0.009 0.006 0.006 0.007 0.008 JS Test 0.289 0.010 0.000 0.086 0.165 0.069 0.149 0.190 Observations 135 135 135 135 135 135 135 135 # Countries 15 15 15 15 15 15 15 15
42 Appendix E. Table 6 – Full regressions
Table 5 a ESI T-1
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.343*** 0.332*** 0.344*** 0.317*** 0.307*** 0.350*** 0.337*** 0.337*** (0.0572) (0.0671) (0.0552) (0.0664) (0.0468) (0.0631) (0.0518) (0.0646) sit 0.241** 0.190* 0.221** 0.211** 0.196* 0.234* 0.238** 0.251** (0.113) (0.103) (0.0863) (0.0994) (0.108) (0.137) (0.0954) (0.121) nit + δ + ga -1.804** -1.594** -1.554** -1.707** -1.305 -1.796* -1.770** -1.990*** (0.886) (0.811) (0.669) (0.799) (0.915) (1.073) (0.826) (0.744) humit -0.0386 -0.0344 -0.0261 -0.0469* -0.0186 -0.0406 -0.0409 -0.0514 (0.0482) (0.0320) (0.0265) (0.0275) (0.0319) (0.0423) (0.0318) (0.0319) Log ESIit-1/GDPit-1 -2.460 0.309 0.200 -0.269 -0.255 -2.040** 0.539 -1.220 (1.710) (0.456) (0.456) (0.606) (1.06) (0.995) (0.90) (0.876) Interaction 0.015 -0.22*** 0.209** 0.198 0.099 2.470*** -0.663 1.120* (0.010) (0.087) (0.083) (0.655) (1.292) (0.947) (0.66) (0.596) AR(1) 0.005 0.003 0.005 0.005 0.007 0.005 0.007 0.007 JS Test 0.333 0.008 0.039 0.901 0.933 0.033 0.567 0.169 Observations 120 120 120 120 120 120 120 120 # Countries 10 10 10 10 10 10 10 10
43 Table 5 b ESI T-2
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.334*** 0.343*** 0.313*** 0.298*** 0.317*** 0.338*** 0.315*** 0.338*** (0.0453) (0.0579) (0.0468) (0.0542) (0.0526) (0.0486) (0.0476) (0.0544) sit 0.237** 0.206* 0.211** 0.228** 0.214* 0.233 0.250** 0.242* (0.117) (0.111) (0.0913) (0.108) (0.126) (0.149) (0.105) (0.124) nit + δ + ga -1.542 -1.425* -1.314* -1.565* -1.238 -1.534 -1.623* -1.655* (1.018) (0.857) (0.699) (0.919) (0.944) (1.200) (0.957) (0.885) humit -0.0355 -0.0343 -0.0298 -0.0577 -0.0364 -0.0418 -0.0539 -0.0473 (0.0482) (0.0339) (0.0242) (0.0362) (0.0344) (0.0449) (0.0407) (0.0338) Log ESIit-2/GDPit-2 -2.240 0.197 0.281 0.013 -0.030 -1.340 0.674 -0.571 (1.740) (0.430) (0.449) (0.626) (1.04) (0.928) (0.874) (1.060) Interaction 0.015 -0.172 0.166* -0.057 0.018 1.830** -0.692 0.586 (0.011) (0.122) (0.096) (0.506) (1.12) (0.884) (0.57) (0.700) AR(1) 0.005 0.004 0.008 0.005 0.006 0.005 0.007 0.007 JS Test 0.393 0.194 0.205 0.993 0.999 0.107 0.453 0.616 Observations 110 110 110 110 110 110 110 110 # Countries 10 10 10 10 10 10 10 10
44 Table 5 c ESI T-3
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.147* 0.127 0.0937 0.147* 0.131* 0.114* 0.144* 0.147* (0.0764) (0.0925) (0.102) (0.0821) (0.0671) (0.0692) (0.0761) (0.0782) sit 0.680*** 0.656*** 0.594*** 0.688*** 0.644*** 0.609*** 0.654*** 0.676*** (0.170) (0.175) (0.178) (0.175) (0.137) (0.155) (0.129) (0.160) nit + δ + ga -4.464*** -4.390*** -4.242*** -4.596*** -4.118*** -3.762*** -4.263*** -4.506*** (1.178) (1.169) (1.068) (1.139) (1.110) (1.334) (1.052) (1.143) humit -0.214** -0.215** -0.219*** -0.219** -0.217** -0.203** -0.206** -0.216** (0.0917) (0.0927) (0.0823) (0.0917) (0.0922) (0.0909) (0.0897) (0.0964) Log ESIit-3/GDPit-3 0.051 0.817** 0.876** 1.230** 1.210 -0.874 1.830* 0.514 (1.310) (0.353) (0.389) (0.592) (1.16) (1.110) (1.08) (1.260) Interaction 0.005 -0.139** 0.056 -0.752 -0.420 2.290** -0.999 0.272 (0.008) (0.065) (0.091) (0.637) (1.26) (1.030) (0.85) (1.180) AR(1) 0.009 0.006 0.010 0.009 0.013 0.007 0.014 0.014 JS Test 0.086 0.000 0.065 0.106 0.046 0.000 0.168 0.136 Observations 100 100 100 100 100 100 100 100 # Countries 10 10 10 10 10 10 10 10
45 Table 5 d ESI T-4
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.166** 0.151** 0.138 0.180*** 0.153*** 0.160** 0.166** 0.169** (0.0737) (0.0754) (0.0921) (0.0662) (0.0572) (0.0657) (0.0721) (0.0774) sit 0.641*** 0.624*** 0.546*** 0.646*** 0.633*** 0.616*** 0.635*** 0.636*** (0.154) (0.165) (0.169) (0.167) (0.157) (0.178) (0.144) (0.162) nit + δ + ga -4.600*** -4.514*** -4.303*** -4.733*** -4.482*** -4.431*** -4.570*** -4.609*** (1.165) (1.100) (1.020) (1.071) (1.147) (1.455) (1.075) (1.127) humit -0.230*** -0.232*** -0.230*** -0.231*** -0.235*** -0.228*** -0.231*** -0.231*** (0.0848) (0.0853) (0.0763) (0.0831) (0.0871) (0.0858) (0.0851) (0.0848) Log ESIit-4/GDPit-4 1.000 2.100*** 1.81*** 2.63*** 2.830* 0.322 3.04** 1.830 (1.570) (0.525) (0.444) (0.747) (1.541) (1.510) (1.256) (1.490) Interaction 0.006 0.208 -0.198 -1.170 -0.926 2.080 -1.220 0.105 (0.010) (0.153) (0.120) (0.792) (1.591) (1.460) (1.012) (1.571) AR(1) 0.006 0.007 0.008 0.006 0.007 0.006 0.005 0.010 JS Test 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.000 Observations 90 90 90 90 90 90 90 90 # Countries 10 10 10 10 10 10 10 10
46 Appendix F. Table 7 – Full regressions
Table 6 a ESI T
(1) (2) (3) (4) (5) (6) (7) (8)
Trade Inflation Budget CoC GE PS RQ VaA
VARIABLES git git git git git git git git
git-1 0.361*** 0.369*** 0.364*** 0.360*** 0.338*** 0.370*** 0.357*** 0.347*** (0.0455) (0.0487) (0.0469) (0.0455) (0.0370) (0.0467) (0.0429) (0.0402) sit 0.179*** 0.103*** 0.139*** 0.129*** 0.153*** 0.200*** 0.152*** 0.175*** (0.0507) (0.0312) (0.0348) (0.0376) (0.0419) (0.0437) (0.0429) (0.0482) nit + δ + ga -1.376*** -1.090*** -1.721*** -1.553*** -1.413*** -1.427*** -1.610*** -1.434*** (0.422) (0.422) (0.445) (0.501) (0.413) (0.429) (0.508) (0.452) humit -0.00950 0.00529 -0.0149 -0.00224 0.00339 0.00882 0.00684 0.0105 (0.0189) (0.0176) (0.0178) (0.0138) (0.0121) (0.0193) (0.0164) (0.0173) Log ESIit/GDPit -3.39*** -0.301 0.590 -0.654 -0.897 -2.87*** -0.809 -1.320 (0.970) (0.633) (0.839) (0.78) (1.11) (0.917) (0.85) (1.030) Interaction 0.019*** -0.173* 0.37*** 0.012 0.377 2.960*** 0.101 0.925 (0.004) (0.102) (0.105) (0.90) (1.28) (0.883) (0.97) (1.170) AR(1) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JS Test 0.000 0.102 0.001 0.513 0.411 0.003 0.267 0.320 Observations 325 325 325 325 325 325 325 325 # Countries 25 25 25 25 25 25 25 25
