Trust in institutions is key to value added tax compliance – global studies prove

6.3. Country-level estimates Our preceding results indicate substantial VAT revenue losses due to non-compliance, especially among countries with low state capacity, as proxied by the WGI index. Those estimates were for groups of countries defined by region or institutional quality, but estimates of compliance and policy gaps for specific countries are of independent interest. Our panels are too short to estimate ci efficiently country-by-country, but heterogeneity in compliance among individual countries remains a subject of considerable interest. The analysis in Section 4 suggests that the Control of Corruption index wi is closely related to compliance gaps. In this section, we therefore estimate a general model of revenue performance Eit ¼ ai þ B wi; Mit Cit þ x0 ith þ uit: ð15Þ using non-parametric methods. The specification (15) contains a large number of regressors, including country and region-by-year fixed effects as well as parametric controls, and non-parametric estimation performs poorly in such settings (Carneiro et al., 2011).35 We therefore estimate the model in two steps. In the first step, we regress Eit and Mit=Cit on the full set of control variables xit and fixed effects, and compute the residuals from each regression. In the second step, we estimate (15) on the residualized Eit and Mit=Cit by local linear regression, using the Epanechnikov kernel and the optimal bandwidth computed using the procedure of Racine and Li (2004). We then recover the marginal effects of the import ratio b ¼ @B=@ð Þ M=C at the means of the data for each country in the sample. Finally, we recover the estimated compliance gap for each country using (9) and the algebraic procedure discussed in Section 3.1. Estimates appear in Table A1 in the Appendix. Fig. 4 graphs the domestic compliance gaps estimated from this approach.36 The estimated compliance gap is broadly decreasing in institutional quality, as expected. The patterns of individual estimates observed in the figure help to elucidate the competing effects of geography and institutions. The highest estimated gaps—in excess of 60 percent of domestic value added—are in African countries, even though not all of them have the lowest values of the governance index. Somewhat lower gaps but still high are observed in a number of Eastern European and (to a lesser degree) Asian countries for example. Fig. 5 shows estimated compliance gaps from the specification with controls for EU countries. This gives a sense of if our measures line up with external measures when the latter are available, and speaks to whether our measures might be useful when they are not. For these countries, accounting estimates of the compliance gap are available using the top down method from Barbone et al. (2013), and subsequent updates by the European Commission. In this case, we report aggregate compliance gaps that adjust for Fig. 3. Institutional quality in the European Union. 34 Of course, this should not be interpreted as a statement about the welfare effects of the policy gap. Selective rate reductions create both winners and losers in the domestic economy and so have ambiguous welfare impacts. In contrast, revenue losses due to the compliance gap have a less ambiguously negative impact on welfare (Emran and Stiglitz, 2005). 35 We employ the same control variables as in previous specifications, as well as region-by-year fixed effects for the seven global regions defined in Table 3. 36 For legibility, the figure excludes the estimate for Republic of Congo, which is 250 percent. P. Morrow, M. Smart and A. Swistak Journal of Public Economics 208 (2022) 104634 11 the higher compliance rate for imports, rather than domestic compliance gaps, since that is what is reported in the previous literature.37 The resulting estimates are therefore considerably smaller. The two sets of estimates are graphed against each other in Fig. 5. Compared to the accounting measure, our estimates are broadly similar. Gaps are largest in the newly acceding countries of Eastern Europe, as well as in Italy and Greece. Compliance gaps are less than twenty percent in many northern and western European countries. Thus, unlike other regions of the world, losses in potential VAT revenues reflect the policy gap far more than the compliance gap in most EU countries. 7. Compliance and institutions: An exploratory analysis The country-level estimates of Section 6.3 suggest a relationship between perceived corruption and VAT compliance at the national level. Cross-sectional patterns depicted in Fig. 4 cannot be taken as evidence of any causal link between compliance and institutional quality among countries, as many other observable and unobservable factors correlated with corruption perceptions may also explain differences in compliance gaps among countries. Moreover, as emphasized by Besley and Persson (2014), political and economic processes of development create a feedback loop between governance quality, tax compliance, and government revenues. In this sense, the measure of institutional quality used in Section 6.3 is just a convenient way to group countries together into those with similar institutions for the purposes of estimation, rather than the demonstration of a causal explanation for the patterns of non-compliance.