Central bank independence
For the government to substitute seigniorage revenues for tax revenues, it must be able to set not only taxes, but also monetary policy. With an independent central bank, the mechanism on which the theoretical model rests breaks down, because the government can simply not control money creation. In other words, the marginal effect of the shadow economy on inflation and taxes should be conditional on central bank independence (CBI). More precisely, we expect the absolute marginal impact of the shadow economy on inflation and tax revenue to be decreasing in CBI. To take that possibility into account, we interacted the size of the shadow economy with a measure of CBI in all our regressions, so as to let the marginal impact of the shadow economy be a linear function of CBI. We used the index of CBI developed by Cukierman et al. (1992) as updated by Crowe and Meade (2008). It is available for two years (1998 and 2006) and 90 countries in our sample.15 This gives us enough variation for pooled regressions. The results are shown in Table 3. The implied marginal impact of the shadow economy and its significance are computed for the minimum, mean, and maximum values of the CBI index, and reported in the
last three rows of the table.
The exchange rate regime
CBI is not the only institutional factor that constrains monetary policy. The adoption of a fix exchange rate regime also takes the reins of monetary policy out of the hands of the government. As a result, the magnitude of the effects that we have so far observed is likely conditional on the exchange rate regime. That magnitude should increase with the flexibility of the regime. To test this contention, we need to interact the size of the shadow economy with the exchange rate regime of the country.
To this end, we employ the exchange rate data set of Ilzetzki, Reinhart, and Rogoff (2008), who classify de facto exchange rate regimes into four categories: pegged exchange rate regimes, crawling pegs, managed floats, and freely floating regimes. We interacted the shadow economy with a dummy variable for each of exchange rate regime, taking fully flexible regimes as the reference category. The implied marginal impact of the shadow economy and its significance for pegged exchange rate regimes, crawling pegs, managed floats are reported in the last three rows of Table 4. The marginal impact of the shadow economy in the reference category is directly given by the coefficient of the shadow economy in each regression.
Developing and developed countries
Our estimations have so far neglected the differences between developed and developing countries. Yet, Gërxhani (2004) emphasizes that the relation between the formal and informal sector likely differ across the two groups. Dreher and Schneider (2010) similarly find that the relation between corruption and the shadow economy differs between developed and developing countries. Although the public finance argument of inflation is not conditioned on the level of development of the country, the structural differences between developed and developing countries may result in the marginal impact of the shadow economy differing across the two groups of countries. We therefore re-estimate the regressions of previous section on two separate samples: one consisting of developing countries, the other including only developed countries. In Table 5a we consider the former while in Table 5b we consider the latter.
By and large, splitting the sample between developed and developing countries does not affect our findings qualitatively. Overall, the models are significant in both sub-samples with the only exception of uni-variate model (5a.1). In both sub-samples, the marginal impact of the shadow economy though differ in magnitude, is positive for inflation, while it is negative in tax equations. Our results are therefore not driven by a particular subset of countries.
Alternative measures of the size of the shadow economy
A possible concern with the size of the shadow economy is that it cannot be directly observed. One may consequently worry about the sensitivity of our results to the specific estimates of the shadow economy we used. Therefore, we verify our results employing alternative estimates of the size of the shadow economy.
Our first alternative is the Johnson et al.’s (1998) estimates. Johnson et al. (1998) provide a single estimate of the size of the shadow economy for 49 countries for various years around 1994. We therefore estimate the relations only with OLS and SUR estimators on a cross-section, using values of inflation and taxes for 1994 in our estimates. Table 10 reports the results using Johnson et al’s estimates of the shadow economy.
As is shown in the lower panel, we have to estimate our model on a sample of less than 40 countries, which does not allow us to tackle all the statistical issues discussed in the previous sections. Nonetheless, in the last two columns we estimate the system of two equations using the SUR estimator to take into account the simultaneity problem as it directly relates to our theoretical model. As shown in the table, all of our models are statistically significant beyond the one-percent level. Remarkably, most of our earlier results hold with this change in the measurement of the shadow economy as well as in the sample size and time period of estimation.
We complement Johnson et al.’s (1998) estimates by those of Elgin and Oztunali (2012), who use a general equilibrium approach. Those estimates are available for a large panel data set, which make them readily substitutable in our empirical models. Therefore, we report, in Table 11, estimates using pooled OLS with panel corrected standard errors and instrumental variable two step GMM estimator. The same estimates are reported for the inflation equation (columns 11.1 to 11.4) and the tax revenue equation (columns 11.5 to 11.8). As can be seen, the results in Table 11 are very similar to their corresponding estimates reported previously. Table 12 addresses simultaneity, and reports the SUR and 3SLS estimates obtained using Elgin and Oztunali’s (2012) estimates. As in the previous case, here too, our results are unchanged.
In summary, the robustness checks bring home the point that our results are due neither to a mis-specified model nor to endogeneity/simultaneity issues. Our results hold both for developing and developed countries, and are equally valid across different measures of the size of the shadow economy.
Transparency and output volatility
For our purposes, we can divide literature on transparency into three broad classes. One dealing with the effects of preference transparency; second focuses on the effects of transparency about the shocks (policy, expectational, or supply shocks). And the third strand relates to the measurement and empirical effects of transparency. Given the primarily empirical nature of our enquiry we restrict ourselves with the third strand of literature. Comprehensive survey of the economic effects of transparency can be found in Blinder et al. (2008), Van der Cruijsen and Eijffinger (2008), and Cukierman (2009).
The empirical verification of theoretical insights requires measurement of transparency. The first comprehensive attempt in this direction is the survey based transparency indicators developed by Fry et al. (2000). Gathering data from 94 central banks, they find that transparency is an important instrument to achieve credibility across policy frameworks. They write ‘our results show that credibility is achieved through discretionary strategy employing a combination of transparency and explanation. Policymakers ask not just to be judged purely on results but instead commit themselves to inform the public about their thought processes so that agents may understand the difficulties of the economic environment’ (p. 139).
However, it is Eijffinger and Geraats (2006) who classify and quantify five aspects of a transparent policy: political, economic, procedural, policy, and operational.26 Unlike Fry et al., they develop time varying index to estimate transparency scores of 9 major OECD central banks from 1998 to 2002. Dincer and Eichengreen (2007) extend the coverage of Eijffinger and Geraats index to 100 central banks with a time period from 1998 to 2006.
The recent updates of these transparency scores by Dincer and Eichengreen (2010) and Siklos (2011) find that transparency is increasing in majority of the central banks. This positive trend in transparency indicates the increasing reliance that central banks are assigning to it in achieving policy objectives.
Endogeneity and Simulteneity
In Table 5 we consider simultaneity and endogeneity issues. Admittedly, it is not easy to determine what comes first: macroeconomic stability or transparency. These issues arise because macroeconomic policies neither focus on one variable nor are they independent of the evolution of these variables. Statistically, therefore, it is possible that our results, rather than reflecting the effect of our regressors, in fact, reflecting the effect of some omitted variable that is correlated with the dependent variable. In that case the causality would be reverse. To take into account this possibility we have estimated our regressions using Arellano and Bond (1991) estimator and system 3SLS model.
Arellano and Bond estimator allows the dynamic specification through the lagged values of the dependent variable.36 Moreover, it uses lagged values of the endogenous variables as their own instruments, an advantageous feature because good instruments are hard to find. We implement this estimator to get results in column (5.1) taking transparency, fiscal balance, and inventory changes as endogenous. The model in column (5.1) corresponds to our baseline model in column (2.1). Because we are considering lagged value of the dependent variable as regressor other control variables are unnecessary. The consistency for Arellano and Bond estimator requires that error term be serially uncorrelated. It can be tested under the null hypothesis that covariance between first differenced error terms is insignificant beyond first order (Cameron and Trivedi, 2005). As is shown in table 5, we cannot reject the null hypothesis of no autocorrelation between first differenced error terms beyond order 1, thus our model qualifies this requirement.
The second assumption for consistency requires the validity of over-identifying restrictions. In other words, the overidentifying restrictions test (or Sargan test) tests the validity of exogenous information that we are using (through instruments) for the identification of causal relation between our variables. In our case, the null hypothesis of validity of over-identifying restrictions is accepted convincingly. Therefore, we can trust our dynamic model. Coming to our results, they are not different from the earlier ones: transparency is having a significant negative effect on output volatility. However, unlike previous estimates, the interpretation here is that transparency has a causal influence on the output variability. By contrast, our results do not suggest causality for other endogenous variables.
Table of contents :
Chapter 1. Introduction
1. An overview of the political economy
2. Political economy of inflation
3. Contribution of the study
4. Research approaches
Chapter 2. Taxing the unobservable: The impact of the shadow economy on inflation and tax revenue
2. A simple theoretical framework
3. Data and econometric methodology
5. Robustness checks
6. Concluding remarks
Appendix I. Tables
Appendix II. Variables definitions and sources
Chapter 3. Transparency and output volatility: International evidence
2. Sources of output stabilization
3. A simple theoretical framework
4. Empirical methodology
6. Robustness analysis
Appendix I. Tables
Appendix II. Variables definitions and sources
Chapter 4. Monetary policy committee transparency: measurement, determinants, and economic effects
2. Aspects of transparency in monetary policy
3. Measuring monetary policy committee transparency
4. Theoretical determinants of MPCT
5. Empirical analyses of determinants of MPCT
6. The effect of monetary policy transparency on inflation variability
Appendix I. Tables
Appendix II. A) Constructing MPCTI
B) Variables definitions and sources
Chapter 5. Conclusion
1. Summary of conclusions
2. Perspective on future research