Inequality, democracy and taxation: Lessons from the post-communist transition

Abstract

Using data for the period 1989 – 2002, we examine the determinants of income inequality in post-communist economies. We find a strong positive association between equality and tax collection but note that this relationship is significantly stronger under authoritarian regimes than under democracies. We also discover that countries introducing sustainable democratic institutions early are characterised by lower inequality. We also confirm that education fosters equality and find that larger countries are prone to higher levels of inequality.

Notes

¹To the extent that they are characterised by greater regularity and transparency, democracies are more amenable to such studies. Indeed, using terminology introduced by Tsebelis (2002): in democracies, ‘veto players’ specified by law (i.e. ‘constitutional’) play a more significant role.

²Following the established tradition, we include the former republics of Yugoslavia and Albania. This yields the set of 27 countries labelled by the European Bank for Reconstruction and Development (EBRD) as ‘transition economies’ (see for example EBRD 2006). For recent discussion, see Mickiewicz (2005) and Havrylyshyn (2006).

³For further discussion see also Aghion et al. (1999).

⁴A related version of this story identifies the role of a narrow class of capitalists at the first phase of the industrialisation process, with a high propensity to save and invest. The resultant inequality emerging during this early stage is seen as a necessary price for economic growth. Kuznets (1963) hints at this interpretation when pointing out that ‘a smaller proportion of the population amasses savings consistently’ in developing countries (Kuznets 1963, p. 48).

⁵Kuznets (1963) chose to emphasise the role of technology and the structure of production although he was fully aware of the interplay between the technological and institutional factors, and between income and wealth distribution and political power relations.

⁶In addition, in transition countries, protracted macroeconomic instability was associated with demonetisation and the collapse of formal incomes. This amplified income inequality (World Bank 2000).

⁷In addition, given the low wages (resulting from the insufficient number of active entrepreneurs), workers may not face incentives to invest in human capital. This, in turn, propagates ‘bad equilibria’ characterised by low education and low wages (Dias & McDermott 2006). One can link Acemoglu (2006) with an earlier seminal paper on regulation of entry by Djankov et al. (2002). In the latter, the elite pursue less efficient policies (via entry barriers) to protect its interests instead of efficient policies (through incentives in the fiscal system which allow for the entrepreneurial sector to grow, in order to tax it later) because their time horizon is shorter (De Long & Shleifer 1993). In contrast, according to Acemoglu (2006) the policy choice is conditional on the overall quality of the fiscal system (i.e. on the level of deadweight cost of fiscal redistribution), where the latter is seen as exogenous with respect to the contemporaneous policy choice. For our purposes, we refer to Section 3 of the Acemoglu (2006) paper. The model is general enough to allow for different interpretations of restricted entry. In particular, it is consistent with an interpretation emphasising the weak protection of property rights, inadequate judiciary systems and corruption, in which only the rich and influential have the resources to protect themselves and small entrepreneurs are discouraged from investment and entry into the formal sector (De Soto 2001, p. 229): ‘Marx would probably be shocked to find how in developing countries much of the teeming mass does not consist of oppressed legal proletarians but of oppressed extralegal small entrepreneurs’. This results in ‘King John's’ redistribution of income rather than ‘Robin Hood's’ redistribution (Glaeser et al. 2003; Sonin 2003; Hellman & Kaufmann 2003). The link between entrepreneurship (self-employment) and exit from poverty is also easily made compatible with Aghion and Commander's (1999) model discussed above. It is sufficient to reinterpret hiring to the ‘new sector’ as a policy parameter affected by entry barriers.

⁸In 2001, Russia introduced a flat income tax of 13%, but the impact of this goes predominantly beyond the time span that our dataset covers. See also Aslund (2002).

⁹The ratio of taxes to GDP decreased in all transition economies temporarily following liberalisation and the initial fall in output (Mickiewicz 2005). However, these effects should be distinguished from the longer-term cross-country variation, which refers to the institutional capacities of the fiscal system as discussed here.

¹⁰However, the sustainability of the current regime in Belarus remains an open question. Its economic stability has relied heavily on subsidised oil and gas provided by Russia (Havrylyshyn 2006).

¹¹Accessed online in January 2006 and used in concert with the user guide produced by the World Institute for Development Economics Research (WIDER) in 2005, available at: http://www.wider.unu.edu/wiid/wiid.htm, accessed 14 January 2006.

¹²There are two reasons for doing this. First, the consumption based measures lead to problems of interpretation when used to assess short and medium-term effects (as in this article). This is because consumption smoothing over time distorts the short-term impact of different economic policies and it is not clear how to value the use of durables in the consumption set. Second, more practically, the comparable consumption-based inequality dataset for transition economies is much smaller, making application of any formal tests very difficult. An example of robust consumption-based inequality data for transition economies is presented in Mitra and Yemtsov (2006). Comparing the datasets suggests that the (positive) gap between income and consumption measures of inequality are greater in less developed transition countries, where consumption data may overstate economic wellbeing at the lower end of the distribution.

¹³Our only exceptions to this second criterion were to retain some estimates for the communist period based on wage earner-only households. We justify this on the basis that comparisons clearly indicate that, given the dominance of this category of income, the results were not sensitive to such a restriction. Moreover, we retained only the cases where such comparisons were possible.

¹⁴The full resulting dataset is available on request from the authors.

¹⁵This was also used in Mickiewicz (2005); see also EBRD (2006). The choice of 1993 split the sample in an even way between ‘early stabilisers’ and ‘late stabilisers’. The ‘early stabilisers’ are Albania, Croatia, the Czech Republic, Estonia, Hungary, Kyrgyz Republic, Latvia, Lithuania, Moldova, Poland, Romania, Serbia, Slovakia and Slovenia.

¹⁶Havrylyshyn (2006) presents further support for this approach. Another interesting set of political indicators can be found in Beck et al. (2001). Unfortunately, it only covers the period up to 1995.

¹⁷Romania switched to full democracy in the late 1990s and Serbia and Montenegro in 2002.

¹⁸We also experimented with resource endowment (which correlates positively with both ‘years under communism’ and size of the country). It did not perform better than the variables we retained in the reported specifications.

¹⁹We also estimate this regression in a fixed effects form replacing the time-invariant variables with country fixed effects. The Hausman test results suggest we may rely on the more efficient random effects estimator; see Table 2 for details.

²⁰More alternative specifications were reported in Gerry and Mickiewicz (2006).

²¹In addition, we also run specifications where the democracy indicators were replaced by the EBRD reform indicator [short run liberalisation indicator constructed as in Falcetti et al. (2002), and early (economic) liberalisation, as in Mickiewicz (2005)], including the interactive term with fiscal revenue. All the results hold and are available on request.

²²Our instrumental variable model is based on overidentifying restrictions (four instruments for two endogenous variables). Correspondingly, we performed another specification test comparing the reported specification with the one using two instruments (i.e. only the export to GDP ratio and its interactive term). We could not reject the more efficient specification, which uses four instruments. It was the latter which was next compared with the model without instrumenting.

²³It should be stressed that for the lagged democracy indicator a negative coefficient implies increased inequality, since higher numbers in the Freedom House index refer to lower levels of democracy, while for the time invariant democracy indicator, a negative coefficient implies decreased inequality.

²⁴Since we consider government expenditure to be constrained by the government's ability to collect taxes, rather than vice versa, we report our specifications with the ratio of revenue to GDP. The results, available on request, using an expenditure measure, do not qualitatively differ. More generally, it would be better to have a direct measure of the quality of the fiscal system. However, such measures are simply not available for our sample. J. Anderson (2005), who focuses on fiscal reforms in transition countries, faces the same problem, and analogous to us uses the ratio of tax revenue to GDP as a proxy for the quality of the fiscal system.

²⁵By way of caveat we note that econometric evidence from earlier studies indicates that stabilisation itself is strongly conditioned by democracy (De Melo et al. 2001; Mickiewicz 2005; Satyanath & Subramanian 2004). Therefore the insignificance identified here may stem from multicollinearity.

²⁶To check the latter hypothesis directly, we run a model for the worldwide sample using the same set of inequality indicators and controlling for a standard set of variables. The communism dummy was insignificant in our estimation (results available on request) and this is robust to variation in specification.

Additional information

Notes on contributors

Tomasz M. Mickiewicz

We are indebted to two anonymous referees, Simon Commander, Phil Hanson, Julia Korosteleva, Branko Milanovic and participants of the seminars at St Anthony's College (Oxford University), City University, London Business School, Brighton University and University of Tartu for valuable comments and criticism.