Freedom through taxation: the effect of fiscal capacity on the rule of law

Abstract

This paper explores the effects of fiscal capacity on the rule of law. We view the question as a natural outgrowth of the stationary bandit model, that rulers are incentivized to make investments in public goods when they are able to extract wealth effectively. We test the relationship using fiscal capacity and rule of law data from the Varieties of Democracy dataset. We leverage the lengthy time-series found in the dataset by employing the dynamic common correlated effects (DCCE) estimator to supplement standard panel methods. Unlike the widely used fixed effects method, DCCE method adjusts for the presence of econometric issues including cross-sectional dependence, heterogeneous slopes, and unobservable common factors that plague the error-structure in panel data. We observe small, positive effects of fiscal capacity on the rule of law, but robustness checks lead us to conclude that our findings, overall, only weakly support the hypothesis.

Appendix A: Robustness checks and incorporation of control variables

It was our starting point that the DCCE method would perform the bulk of the needed identification for what our analysis required. However, we augment our baseline specifications, first, with three control variables, the natural log of real GDP per capita,¹² the clean elections index (v2xel_frefair), and an index of civil society (v2csprtcpt) from Varieties of Democracy. We use the clean elections index, rather than a broader measure of democracy, out of concern that broader measures of democracy incorporate variables that are too conceptually similar to the rule of law, rather than democratic political institutions.

The clean elections index seeks to answer, “to what extent elections are free and fair?” by aggregating eight sub-indicators from Varieties of Democracy: elections management body autonomy, elections management body capacity, election voter registry, election vote buying, election other voting irregularities, election government intimidation, election other electoral violence, and election free and fair. Also from Varieties of Democracy, the index of civil society is derived from country expert answers to the question “Which of these best describes the involvement of people in civil society organizations (CSOs)?”. Ordinal responses are range from, “Most associations are state-sponsored, and although a large number of people may be active in them, their participation is not purely voluntary” to “There are many diverse CSOs and it is considered normal for people to be at least occasionally active in at least one of them.”

Table 6 presents results for eight specifications, the first four of which use the sample of countries beginning in 1880, and the second with the sample of countries beginning in 1950. We run through our main DCCE specification for each of the samples (four lags for the first sample and three for the second) along with each of the control variables alone, and then all of them together. Of the eight specifications, two were statistically insignificant, three were significant at the 0.10 level, and three were significant at the 0.05 level.

Table 6 Results using control variables

We then turn to variables relating to the “durability” of a regime. In Olson’s model of institutional development, the time-horizon of the ruler determines whether the ruler has an incentive to invest in public goods such as fiscal capacity and rule of law. To account for this aspect of Olson’s model, we add control variables to the baseline model that attempt to measure the time-horizon of the leader. These variables include the number of years since a regime change (“durable”), the years the chief executive has been in office, and the shortest tenured veto player. The first two columns control for the “durable” variable, which is a count variable of the number of years since a regime change from the Polity 5 dataset. A regime change is defined as a three-point change in the polity measure of democracy and autocracy.

Since this conception of durability does not strongly relate to individual incentives, we also explore a few alternatives. The estimates in the remaining four columns control for variables from the Database on Political Institutions (data first available in 1975). Columns 3 and 4 control for the years the chief executive has been in office and, columns 5 and 6 control for the shortest tenured veto player. Counter to Olson’s model, these variables enter the regression with a negative sign, indicating that longer tenure is associated with less rule of law. These results suggest that, for many observations, the tenure variables are likely capturing the degree autocratic power or dysfunctional political institutions, with large values corresponding to long tenured dictators. If this is the case, the tenure variables may be acting as collider variables with fiscal capacity that sap the explanatory power of fiscal capacity, our main variable of interest (Table 7).

Table 7 Controlling for measures of tenure

Finally, the estimates in Table 8 use an alternative measure of the rule of law. Instead of the more encompassing rule of law index from V-Dem applied in the main text, these specifications apply the executive constraints measure from the Polity 5 dataset. The variable measures “the extent of institutionalized constraints on the decision making powers of the chief executive” (Marshall, 2020: 23). Using this alternative measure has two drawbacks. First, it is taking a much narrower conception of the rule of law than the V-Dem rule of law index; it is laden with the assumption that the rule of law is necessarily inherent in the structure of political institutions.

Table 8 Alternative measure of the rule of law—executive constraints polity 5 (1965 to 2018)

Second, the measure from Polity 5 has much less data coverage in practice than V-Dem, not only because some countries do not receive a score until the 1960s, but just as importantly, many must be dropped when applying the DCCE methodology because there is no within-country variation in the executive constraint data over the sample period.¹³ If we simply use 1950 as the initial year, the sample of countries drops from 100 to 31. We therefore chose to shorten the time dimension for this robustness check to the period 1965 to 2018, allowing for 61 countries to remain in the sample. In these specifications, fiscal capacity generally has a positive but statistically insignificant association with the rule of law.

As one final robustness check, we use contract-intensive money as another measure of the rule of law (Clague et al., 1999) with a very long time component.¹⁴ Table 9, structured identically to Table 8, reports the results. Here, fiscal capacity even more weakly predicts the rule of law. Although we may protest that contract-intensive money is an indirect indicator that is far weaker than V-Dem, it nonetheless demonstrates that our findings above are sensitive to how rule of law is defined.

Table 9 Alternative measure of the rule of law–contract intensive money (1965–2008)

Results which straddle the edge of statistical significance should be interpreted with a great deal of caution (Ritchie, 2020: 85–104). We can give certain caveats concerning the validity of including variables in these tests (namely that they can be interpreted as collider variables if any of these are the channels through which fiscal capacity impacts the rule of law), or whether the alternative measures of the rule of law are as effective as the V-Dem data, but they lead us to conclude that our findings, overall, only weakly support the original hypothesis. Hence our final assessment found in the conclusion is somewhat subdued.