Keywords

The previous chapter examined the formation of equality preferences. This chapter shifts the focus to policy formulation, implementation, and their outcomes. Most previous studies on the determinants of income inequality across countries included emerging and advanced democracies in one sample, which raises concerns that the independent variables’ effects might be largely attributed to the differences between the emerging and advanced democracies. This study mainly addresses variations among emerging democracies while also outlining the differences between emerging and advanced democracies. Thus, it investigates the effect of political market failure and weak state capacity upon income inequality using an unbalanced panel dataset for the 1985–2012 period for emerging democracies across continents (N = 57), advanced democracies (N = 18), and all democracies (N = 75). The fixed effects (FE) model with a lagged dependent variable (LDV) was adopted as it controls for the country-specific effects, captures the gradual nature of change in income distribution, and errs on the conservative (underestimate) side in coefficient estimation. The dependent variable is the estimated after-tax Gini coefficient. The challenge of finding measurements of incremental and cumulative change in institutional quality was addressed by choosing (1) the age of the largest opposition party for political market and (2) higher-order lags of the Quality of Government indicator and the control of corruption for state capacity . The analysis yielded strong evidence that political market quality and state capacity reduce income inequality; however, the latter takes more time to show its effect. Robustness checks for influential observations (regions and time periods) and an alternative dependent variable (the poorest 20 % of the population) supported these results.

5.1 Conceptualization: Political Market and State Capacity

5.1.1 Political Market: Programmatic Competition and Political Kuznets Curve

As Chap. 2 elaborated, political market quality depends on the clarity of party orientation and programmatic party competition . A lack of clarity in party orientation and programmatic competition implies that voters choose parties not as an indication of policy preferences but due to (provided or expected) patronage (Hagopian 2009). Although the dataset complied by Kitschelt (2014) includes variables for clientelistic and programmatic tendencies, those variables pertain to a single time point (2008 or 2009) and are thus time-invariant. The FE model adopted in the current analysis can incorporate only those variables that change over time. Among possible time-variant variables, the mean age of political parties is often used to measure party system institutionalization (Hanusch and Keefer 2014; Gehlbach and Keefer 2012).Footnote 1 This measurement, however, comprises two components that have contrasting effects on programmatic party competition . On the one hand, the age of the governing party may mirror the lack of competition in predominant party systems—a situation that often characterizes emerging democracies (Mozaffar and Scarritt 2005; Doorenspleet and Nijzink 2013)—whereas in competitive party systems, the age of the governing party may simply reflect maturity. The general effect of the governing party’s age on party competition for all types of party systems can be either weakly negative or mixed. On the other hand, opposition parties are particularly ephemeral in those predominant party systems; most of them gradually disappear every election (Mozaffar and Scarritt 2005). In competitive party systems, too, the lack of left–right (programmatic) party competition gives birth to new radical opposition parties that channel voter grievances; the presence of established opposition parties is a strong indication of programmatic competition (Roberts 2013).

The age of major opposition parties—in particular, that of the largest opposition party —thus better captures the level of programmatic competition than that of the governing party or the mean age of political parties. The largest opposition party represents the core element of the opposition and the strongest challenge to the incumbent in programmatic competition . Using the largest opposition party’s age avoids conflating one-party dominance with political market quality. This variable is valid even when the largest opposition party is a former regime party. Former regime parties are generally well organized and well-known by the voters; new incumbent parties therefore face serious electoral challenge to present a clear policy to the electorate (Croissant and Völkel 2012; Hicken and Kuhonta 2011; Smith 2005).

In a broader context of political market quality, some scholars suggest that the distributive effect of democratization changes over time. According to the political Kuznets curve theory , democratization initially expands the income gap between the rich and poor but eventually narrows it (Acemoglu and Robinson 2002; Chong 2004; Tam 2008). This argument, primarily based on Western European history prior to the early 20th century, highlights the impact of a gradual expansion of suffrage and education. Their expansion initially increases inequality because only the elite benefit from them; however, later, universal suffrage and mass education help to reduce inequality (Acemoglu and Robinson 2002; Bourguignon and Verdier 2000). Supportive findings were obtained from analyses of panel datasets that include both democratic and non-democratic countries covering the period from the 1960s to the mid-1990s (Chong 2004; Tam 2008).

However, in almost all emerging democracies and their non-democratic predecessors analyzed in this study for the 1985–2012 period, universal suffrage has been established. Geddes (2007) also highlights that numerous non-democratic regimes in the late 20th century pursued redistribution at the expense of traditional elites through land reform, expanded education, and industrialization. Moreover, compared with the datasets used by Chong (2004) and Tam (2008), the panel dataset used in this chapter encompasses longer periods of democracy than non-democracy. The poor were underrepresented because of the lack of parties that represent their interests under non-democratic systems. In this sense, democratization provides the electorate a greater choice of representatives. Thus, greater competition for public office realized by democratization does not privilege the rich as in the case of suffrage expansion; on the contrary, it favors the poor over the rich who enjoyed easier access to state authorities than the poor when electoral competition was restricted. Although the effect of democratization is expected to benefit the poor immediately rather than later, this study’s analysis tests for the presence of a political Kuznets curve.

5.1.2 State Capacity: Corruption’s Kuznets Curve

The effect of state capacity on income equality hinges substantially on controlling corruption . Corruption has been considered to aggravate inequality by increasing tax evasion, thus benefiting the rich while also reducing social expenditures designed to assist the poor. This claim, however, is supported only by cross-sectional studies (Gupta et al. 2002; Gyimah-Brempong and de Gyimah-Brempong 2006) or a panel analysis of 10-year interval first-differentiated data (Chong and Gradstein 2007). Panel analyses of data with time intervals of four or fewer years do not report such a monotone relationship (Andres and Ramlogan-Dobson 2011; Dobson and Ramlogan-Dobson 2012). Even among cross-sectional studies, Chong and Calderón (2000) demonstrate that institutional quality has an inverted-U curve effect on inequality, while deriving the contrasting conclusion that institutional quality increases equality in developed countries but reduces equality in developing countries. We argue that these puzzling results arise because, in the short term, the spurious positive effect of corruption on inequality overwhelms the genuine negative effect of corruption.

The spurious positive effect of corruption emerges for two reasons. First, redistributive policies aimed at reducing inequality inherently foment corruption; greater corruption thus appears to reduce inequality while, in reality, redistribution is the actual cause. Alesina and Angeletos (2005) indicate that large governments and redistributive policies induce corruption through tax loopholes, project allocations, and regulations that favor rent seekers. Second, policy measures to rein in corruption affect the poor by reducing the informal sector that most rely on to generate their incomes (Andres and Ramlogan-Dobson 2011; Dobson and Ramlogan-Dobson 2012; Balafoutas 2011).Footnote 2 de Freitas (2012) argues that a large informal economy, by evading income taxes, induces the government to resort to indirect taxes, which are theoretically more regressive than income taxes. In the same context, Mahon (2011) shows that tax reforms in Latin America increased inequality, apparently because they relied heavily on indirect taxes. Changes in corruption may thus reflect policies for reducing inequality or those for controlling corruption. While a reduction in corruption may contribute to income equality in the long run, policies that reduce corruption (or inequality) may increase inequality (or corruption), at least in the short term.

The current analysis tests our claims that (1) the contemporaneous effect of state capacity (including corruption ) on income inequality is more spurious than real and (2) the long-term effect of state capacity on income inequality is negative. Although the negative effect of corruption on inequality was evidenced thus far only by cross-sectional studies or a panel analysis of very long-interval (10-year mean) data, we replicate that effect using an annual panel that includes levels of state capacity going more than four years back. The effect of state capacity reverses in its higher-order lags. The complex effect of state capacity is thus scrutinized in both the short and long term.

5.2 Research Design

5.2.1 Data and Samples

This study separately tests the effect of the three factors on income inequality for emerging democracies across continents (N = 57), advanced democracies (N = 18), and all democracies (N = 75), using an unbalanced panel dataset for the 1985–2012 period compiled from the Standardized World Income Inequality Database (SWIID), the International Country Risk Guide (ICRG), the Quality of Government Database, the World Development Indicators Database, and other sources. Democracies were defined as countries whose Polity2 score was at least 6 (in accordance with the definition of democracy in Polity IV) for any four consecutive years, which usually forms one presidential or legislative term, during the 2001–2012 period. This definition encompasses all democracies that have existed in the 21st century, including those that became a democracy or reverted to a non-democracy during the 12-year period. The democracies were then divided into (1) emerging democracies that became either independent after 1944 or democratic after 1959, and (2) advanced democracies that were both independent before 1945 and democratic prior to 1960. As exceptions to this definition, Colombia and Costa Rica were categorized as emerging democracies. See Table 5.1 for the sample of countries.

Table 5.1 Number of observations by country for model 1 for all democracies in Table 5.5 (N = 1275)
Full size table

5.2.2 Panel Design

The panel analysis adopted the FE model with an LDV due to its better match with the current dataset in comparison with other models. Alternatives to the FE model such as a random effects model or a panel corrected standard errors (PCSE) estimation did not meet the dataset property. A random effects model was not chosen because the Breusch-Pagan test, by rejecting the null hypothesis of no dependence of variance on country, indicated that the independent variables were correlated with unobserved country effects. PCSE estimation is appropriate for a panel with a limited number of cross-sections for a long time period but not for a panel having more cross-sections than time points (Beck and Katz 1995), which is the case here. Hence, PCSE estimation was also rejected as an approach.

The FE model mitigates a potential problem of selection bias arising from unbalanced panels (or different numbers of observations per country) such as this dataset, because the country-specific intercept, which represents unobserved effects, captures the idiosyncratic likelihood of absent observations (Wooldridge 2013, pp. 473–74). Furthermore, the FE model can accommodate an LDV model, which has three appealing properties in the context of the current research. First, the model is appropriate for situations where the effect of a change in an independent variable is distributed over time. Second, although the inclusion of an LDV makes the FE (and OLS) estimator inconsistent, the FE (not OLS) estimator becomes consistent when T becomes large. An appropriate value for T is 20 or greater according to Beck and Katz (2011, p. 342) while Baltagi (2008, p.148) cites an example of relatively consistent estimators when T reaches 30.Footnote 3 As the mean observation per country in the dataset is 14.9 for the emerging democracies and 16.9 for all democracies, potential estimator inconsistency should be far from serious.Footnote 4 Third, misspecification in the LDV model would lead to underestimation rather than overestimation of regression coefficients (Beck and Katz 2011, p. 336).Footnote 5 This tendency for underestimation prevents us from erroneously asserting significant impacts of the variables of interests.

In sum, the FE model with a LDV has three major advantages over other models. First, it enables addressing the question of whether socioeconomic and political changes account for incremental change in each country’s income distribution. Second, it controls for country-specific conditions such as colonial experiences and path dependence more generally; it also reduces the selection bias inherent in unbalanced panels. These features of the model well serve the major interest of this study, which is to determine the impact of political and economic reform on income equality in emerging democracies and not to undertake a comparison of income equality among countries at different levels of democracy. Third, conservative estimates of variable coefficients deter a false claim of new evidence. Such caution is all the more necessary when the operational hypotheses rest on less-than-solid ground. The FE model with a LDV used here takes the following form:

$$D{V_{i,t}} = \alpha + {\beta _1}(D{V_{i,t - 1}}) + {\beta _2}(IV{1_{i,t - 1}}) + {\beta _3}(IV{2_{i,t - 1}}) + \ldots + {\beta _k}(IV{h_{i,t - 1}}) + {\upsilon _i} + {\gamma _t} + {\varepsilon _{i,t}}$$

where DV i,t is a measure of the dependent variable in country i in year t, IV1, IV2, … IVh with h independent variables, α is the intercept, β k are k coefficients to be estimated, υ i are fixed group effects, γ t are fixed time effects, and ε i,t is a white-noise error term.

5.2.3 Variables

Table 5.2 presents the variables and their data sources. The variables of interest are political variables whereas control variables comprise economic, demographic, and year or group dummy variables. All independent variables (variables of interest and control variables) were lagged by one year in the standard specification of the model. The variables for which there are concerns about endogeneity—such as the Quality of Government (QOG) indicator, corruption, and logged GDP per capita—were lagged by more years in extended models (see below).

Table 5.2 Variables and data sources
Full size table

5.2.3.1 Dependent Variable

The dependent variable, the after-tax Gini coefficient , is derived from the SWIID compiled by Solt (2009), who estimated before-tax (“market”) and after-tax (“net”) Gini coefficients as well as changes in the Gini coefficient after taxation (“redistribution”) using the World Income Inequality Database (UNU-WIDER 2008), the Luxemburg Income Study Database (LIS 2008), and more recent country-specific databases. In this study, the estimated before-tax Gini coefficient and the estimated redistribution were also used as alternative dependent variables; however, the estimated after-tax Gini coefficient returned the most substantive results. As a robustness check, the income share held by the lowest 20 % of the population was used as an alternative dependent variable.

5.2.3.2 Political Market

The quality of political market was measured by the age of the largest opposition party (Beck et al. 2001). The relative validity of the largest opposition party variable in comparison with alternative party age variables, such as the mean party age, the age of the executive party, or the age of the largest government party, can be checked by examining whether the relevant party age is associated with economic policy competition between the incumbent and opposition. Economic policy competition was measured by the (legislative) polarization variable (Beck et al. 2001). Despite its connotation, the polarization variable indicates whether party competition in the legislature is either left versus right (=2), center versus left or center versus right (=1), or no programmatic competition (=0).Footnote 6 The correlation Table 5.3 demonstrates that the age of the largest opposition party is more strongly associated with polarization than any other party age variable regardless of logarithmic transformation or recoding of those variables. The alternative party age variables as well as corresponding party seat variables were also used for preliminary panel analyses but none of them had a significant effect on income inequality.

Table 5.3 Pearson correlation coefficients for party age variables and legislative polarization, all democracies (N = 1912)
Full size table

As an alternative measurement of political market quality, the Freedom House/Imputed Polity2 variable in the Quality of Government Database (Teorell et al. 2013) was adopted, which is calculated as a composite indicator of the Freedom House score and Polity score. Freedom House uses minority rights as one criteria when calculating its score. The question on its checklist most relevant to minority rights asks, “Do cultural, ethnic, religious, or other minority groups have full political rights and electoral opportunities?”Footnote 7 In contrast, the Polity score focuses on checks and balances in political institutions but does not explicitly specify any element of minority rights. This composite variable thus captures political competition and minority representation in a balanced way. Although this variable is less focused on the level of programmatic party competition compared with the age of the largest opposition party, the fact that it comprehensively measures political competition among political parties and groups allows for testing the political Kuznets curve hypothesis that democratization initially increases income inequality because only the privileged enjoy political participation at its early stage (Acemoglu and Robinson 2002).

5.2.3.3 State Capacity

The effect of state capacity was measured by the Quality of Government (QOG) indicator (Teorell et al. 2013), calculated from three variables included in the International Country Risk Guide (ICRG) dataset (PRS 2013)—control of corruption (a ratio scale ranging from one to six), the rule of law (a ratio scale ranging from one to six), and bureaucratic quality (a ratio scale ranging from one to four)—and standardized to range between 0 and 1. The variables in the ICRG dataset are compiled by the Political Risk Service (PRS) using specialist evaluations of various political and economic risks of the countries around the world.Footnote 8 Among the three variables that form the QOG indicator, the control of corruption can have the most influential effect. QOG was thus replaced with the control of corruption per se in alternative models. The contemporaneous effect of state capacity was measured by the first lag of QOG or the control of corruption. Its long-term effect was gauged using different higher-order lags as well as means for five consecutive higher-order lags of QOG or the control of corruption.

5.2.3.4 Ethnic Peace

This study assumes that multidimensionality is a more proximate cause of policy input (preference formation) than that of policy outcome (redistribution or inequality reduction). The previous chapter demonstrated that the multidimensionality of policy issues, operationalized by ethnic fractionalization, discourages the formation of preferences for income equality. However, it does not preclude ethnic fractionalization from affecting income equality (1) directly or (2) indirectly through preferences. There is cross-national evidence that ethnic fractionalization negatively affects redistribution or income equality (Alesina and Glaeser 2005, pp.140–43; Huber and Stephens 2012, p. 145). While it is necessary to examine the (direct or indirect) effect of ethnic fractionalization on income equality, the fixed effects model adopted in this chapter cannot accommodate time-invariant variables such as ethnic fractionalization.

Instead, the following analysis redirects the focus onto an activation of multidimensionality . An activation of multidimensionality, measured by ethnic tensions, may exacerbate income inequality by facilitating ethnic-based coalitions rather than lower-income coalitions. Two caveats must be highlighted. First, although an activation of multidimensionality is not independent of ethnic fractionalization (because societies that are purely homogeneous in terms of ethnic groups cannot have ethnic tensions), it cannot be assumed that ethnically more heterogeneous countries trend to have greater ethnic tensions. Second, an activation of multidimensionality in ethnically heterogeneous countries may have a different effect on income inequality than in ethnically homogeneous counties. The analysis of activated multidimensionality is thus not a substitute for the analysis of multidimensionality per se; it involves more uncertainties and is more explorative than the latter. The variable that measures the absence of ethnic tensions is available at the PRS (2013). This variable, renamed in the current study as ethnic peace , measures tensions in a country that arise from racial, ethnic, or linguistic differences at a ratio scale ranging from one (the highest level of ethnic tensions) to six (the lowest level) in the same manner as used for the control of corruption variable.

5.2.3.5 Control Variables

Control variables were chosen in accordance with the literature (see Table 5.4). The following variables were used as correlates of income inequality (expected effects shown in parentheses): the logarithm of real GDP per capita (+) and its square (−), inflation (+), secondary school enrollment (−), the young population (−), the old population (+), the urban population (−), trade openness (±), and foreign capital investment (±).Footnote 9 Year dummies control for concurrent shocks (e.g., a world economic crisis) and time trends (e.g., neo-liberalism). Kuznets (1955) argued that economic development has an inverted-U curve effect on income inequality but there have been few panel studies to support his theory; most of the supporting evidence is derived from cross-section studies that are prone to unobserved country-specific effects [see the review by Tam (2008)].Footnote 10 Among the control variables, young and old age groups in the population had inconsistent estimates for different models within the same sample. Therefore, the two variables were dropped from the final models shown in the results section. The data source for these control variables is the World Development Indicator Dataset.

Table 5.4 Cross-section time-series studies on democracy, social spending, and income distribution in developing countries
Full size table

Previous studies also included variables related to social expenditures but these were not used in the current analysis for the following reasons. First, although social spending data are available from the IMF’s Government Financial Statistics (GFS), significant discrepancies exist in the GFS data compiled before and after 1990. Specifically, in the post-1990 dataset, two different values were recorded to reflect both accrual and cash basis accounting.Footnote 11 One problem is that most records on cash-based activities do not include non-monetary flows, whereas those on the accrual basis include both monetary and non-monetary flows. For each country, data entries do not necessarily follow the accrual-based system, the cash-based system, or either in a consistent manner. Second, the above dataset has a large number of missing values for emerging democracies, which would significantly reduce their sample size. Third, previous studies indicate that democratic developing countries have higher social spending than non-democratic developing countries (Kaufman and Segura-Ubiergo 2001; Rudra 2002; Avelino et al. 2005) but income equality does not significantly differ between the two groups of countries (Lake and Baum 2001; Rudra 2004; Huber et al. 2006; Ross 2006; Lee et al. 2007). In particular, Huber and Stephens (2012) showed that social spending had no significant effect on income equality in Latin American countries. Therefore, it was judged more reasonable to drop social spending variables and retain the current sample size than to include them and reduce the sample size.

5.3 Results

The results of multiple imputations using the FE model with an LDV are presented in Table 5.5. Models 1 and 2 estimated short-term effects of the independent variables that were lagged by one year. Models 3 through 6 examined the long-term effect of state capacity. The six models were run for the three samples, namely, all democracies, emerging democracies, and advanced democracies. This section concentrates on results for emerging democracies while referring to the differences from the other two samples. The last two models for advanced democracies could not be estimated because the set of omitted variables or categories was inconsistent for some imputations. This problem emerges when multiple imputations are applied to a small sample.

Table 5.5 Estimation results
Full size table

5.3.1 Political Market

The age of the largest opposition party —as the political market variable—has a negative effect on income inequality for the all democracies or emerging democracies groups at the 0.05 or 0.10 significance level depending on the model but not in advanced democracies as their political market quality is invariably very high. The Freedom House/Imputed Polity2 variable also has a negative effect on inequality in all or emerging democracies. Simultaneously, the effect of the Freedom House/Imputed Polity2 squared variable is positive. This indicates that democratization reduces inequality, at least at its initial phase. This difference from the earlier findings on the political Kuznets curve might be explained by the greater proportion of non-democracy observations in the previous studies than in this study. Furthermore, for the current sample, the effect of political market quality on inequality becomes positive only at the highest level that might be associated with growing income inequality in advanced democracies. In emerging democracies, therefore, even though democracy is immature, enhancing political competition and minority rights ensures greater income equality (Table 5.6).

Table 5.6 Estimation results with an alternative political market quality measurement: Freedom House/Imputed Polity2
Full size table

5.3.2 State Capacity

The effect of state capacity , when measured by the QOG variable, on income inequality has an inverted-U curve, a finding that reflects the fact that the QOG variable and its square have positive and negative effects, respectively, as shown in Model 1.Footnote 12 The inverted-U curve is still present when the QOG variable is replaced with the control of corruption variable (Model 2) (or with Transparency International’s Corruption Perception Index, the results of which are not shown in this study). The inverted-U curve effect of state capacity on inequality was reported by Chong and Calderón (2000) for cross-section data, but this study corroborates these earlier findings with a panel data. These findings are congruent with the argument made earlier in this study that the control of corruption spuriously increases inequality in the short term.

In the long term, however, the quality of government (or control of corruption) contributes to income equality. Models 3 through 6 demonstrate that the quality of government (or control of corruption) as its eighth lag, or as the mean for its sixth to 10th consecutive lags, reduces inequality.Footnote 13 The negative effect of the quality of government (as well as control of corruption) on inequality was the strongest in its eighth lag, and statistical significance declined through the seventh to sixth lag on the one hand, and through the ninth to 10th lag on the other, until the effect became not statistically significant by the fifth and 11th lags. The effect of corruption on inequality became positive by the third lag. The effects of the second and first lags were stronger than those of the third but were very similar to each other. Such a lag effect was not observed for per capita GDP.

5.3.3 Ethnic Peace

The variable for ethnic peace is correctly signed except for two models for emerging democracies, but was only statistically significant in two of the four models for advanced democracies at the 0.10 level. These findings suggest that the activation of multidimensionality is more likely to increase rather than reduce inequality but the effect is far from substantive. The same analysis with split samples of more- and less-fragmented countries yielded results similar to those from the full sample for all, emerging, and advanced democracies. One might speculate that low multidimensionality does not necessarily alleviate the negative effect of ethnic tensions on the formation of a redistributive coalition.

5.3.4 Control Variables

Most control variables have the expected signs, though none are significant for all models. Three other findings are worthy of note. First, trade openness had insignificant but consistently negative signs for the “all democracies” and “emerging democracies” groups and insignificant but consistently positive signs for the “advanced democracies” group. These contrasting results indicate the possibility that trade liberalization in developing countries benefits lower-skilled workers and labor-intensive sectors of the economy while the major beneficiaries are higher-skilled workers and capital-intensive sectors in developed countries. Second, foreign direct investment consistently has positive signs in all samples, although they are only significant in a few models. Third, the logarithm of the GDP per capita and its square, although consistently correctly signed, are not statistically significant for most models. In the current context, income’s Kuznets curve effect is absorbed by corruption’s Kuznets curve. Models 1 and 2, when run without the QOG variable and the control of corruption variable, respectively, returned statistically significant estimates of the logarithm of the GDP per capita and its square (not shown in this study but available from the authors upon request).

5.3.5 Robustness Check

The robustness of the above findings was examined for the sample of emerging democracies in two ways. First, we searched for influential observations by rerunning the most parsimonious and fittest model, Models 3, with one region of countries at a time dropped from the sample for a total six regions (East Asia and Pacific, South Asia, Middle East, Sub-Saharan Africa, Eastern Europe and the former Soviet Union, and Latin America and Caribbean). Although the fixed effects model controls for country-specific effects, certain independent variables might exert particularly strong effects in some countries but only weak effects in others. The estimation results for the six rounds presented in Table 5.7 shows, however, that both the age of the largest opposition party and QOG were statistically significant. In other words, regardless of the region of the world, political market quality and state capacity help to reduce income inequality. The same model was also tested for two shorter time periods, namely 1991–2012 and 1996–2012. For both periods, the age of the largest opposition party (p = 0.030 and p = 0.069, respectively) and QOG (p = 0.002 and p = 0.002, respectively) were statistically significant.

Table 5.7 Robustness check for influential observations: one region dropped from the sample of emerging democracies
Full size table

Second, since the SWIID is based on the standardization of various types of Gini coefficients, the most common alternative measurement of income inequality, i.e., the income share held by the lowest 20 % of the population, was used to check the robustness of the above findings. The country and year coverage of these data are much smaller than that of the SWIID. They do not include advanced democracies and the number of emerging democracies had to be reduced to 26, less than half the original size. The six models for emerging democracies in Table 5.5 were replicated with the lowest 20 % income share as the dependent variable.Footnote 14 The results shown in Table 5.8 reveal remarkable similarities with the earlier results regarding the effect of political market quality and state capacity. For political market quality, the age of the largest opposition party is correctly signed and significant except for Model 4. Freedom House/Imputed Polity2 is correctly signed although not significant for any model. For state capacity, both QOG and control of corruption in their 8th lag or their means for the 6th–10th lags were significant. Although the results for socioeconomic control variables were less consistent throughout the six models, the estimates for the two political variables of interest, i.e., market quality and state capacity, thus give strong support for the earlier findings presented in Table 5.5.

Table 5.8 Robustness check with an alternative dependent variable: income share held by the lowest 20 % of the population
Full size table

In conclusion, the evidence presented in this chapter supports the hypotheses that political market failure and weak state capacity increase income inequality in emerging democracies; however, the activation of multidimensionality does not significantly affect inequality in emerging democracies although it partly accounts for variations in inequality for advanced democracies. In sum, the main political reason for the failure of emerging democracies in improving income equality lies in the lack of party system institutionalization and governance reform. Ephemeral opposition parties are more likely to be personalistic or catch-all than programmatic and thus fail to generate policy competition with the incumbent. Anti-corruption policies exert ambivalent effects on equality in the short term; the merit of enhanced governance takes time to materialize as greater levels of equality.