Two dimensions of democracy and the economy

Abstract

Does democracy influence economic policymaking and outcomes? Our study investigates the implications of Dahl's two dimensions of democracy (‘polyarchy’): contestation/competition and inclusion/participation. We hypothesize that increases in democratic competition inspire policy incrementalism, thus lowering growth volatility and generating fewer deep crises. Meanwhile, increases in substantive democratic inclusion – genuine political voice, or democratic participation in the presence of a minimum of contestation – should increase the political weight of relatively poor voters, who have a differentially strong aversion to deep growth crises. A statistical analysis of 149 countries for 1961–98 finds greater democracy associated with fewer years of sharply negative growth (‘crisis’), with both democratic contestation and substantive inclusion contributing to this outcome. Our conclusions question the wisdom of designing economic policy institutions that are intentionally insulated from the democratic process.

Keywords:

• Democracy
• growth
• economic crisis
• inclusion
• political economy

Acknowledgements

The authors acknowledge Branimira Slavova, who developed the project's early data sets and econometric work, and Reed College, which provided initial funding. We also thank Michael Coppedge, Juliano Assunção, Michael Berbaum, Thomas Borcherding, Tim Carter, John Echeverri-Gent, Robert Franzese, John Gerring, Robin Grier, Charles Hankla, John Jackson, A.M. Mobarak, David Nickerson, Sybil Rhodes, Mark Setzler, Gregory Wawro, Kurt Weyland, Richard Williams, and two anonymous reviewers for helpful comments. Any remaining errors are ours.

Notes

Dahl, Shapiro, and Cheibub, Democracy Sourcebook, especially sections on ‘Defining Democracy’, 1–54, and ‘Democracy and Constitutionalism’, 191–256.

Acemoglu and Robinson, ‘Theory of Political Transitions’; Boix, Democracy and Redistribution.

Cf. Schmitter and Karl, ‘What Democracy Is’; Huntington, Third Wave; Collier and Levitsky, ‘Democracy with Adjectives’; Zakaria, ‘Illiberal Democracy’; Diamond, Developing Democracy.

Dahl, Polyarchy; Dahl, Democracy and Its Critics.

Coppedge, Alvarez, and Maldonado, ‘Two Persistent Dimensions’.

Sen, Development as Freedom.

Rodrik, ‘Where Did the Growth Go?’; Hamilton, Madison, and Jay, Federalist Papers; Brown, Lynn-Jones, and Miller, Debating the Democratic Peace.

Huntington, Third Wave.

Barro and Sala-I-Martin, Economic Growth; Easterly, Elusive Quest for Economic Growth.

Przeworski and Limongi, ‘Political Regimes and Economic Growth’; Brunetti, ‘Political Variables’; Sirowy and Inkeles, ‘Effects of Democracy’.

Przeworski et al., ‘Democracy and Development’, 178–9.

Doucouliagos and Ulubasoglu, ‘Democracy and Economic Growth’, 61.

Quinn and Wooley, ‘Democracy and National Economic Performance’; Almeida and Ferreira, ‘Democracy and the Variability of Economic Performance’; Mobarak, ‘Democracy, Volatility, and Economic Development’; Chandra and Rudra, ‘Regime Type and Economic Performance’.

Przeworski et al., ‘Democracy and Development’, 176–8; cf. Almeida and Ferreira, ‘Democracy and the Variability of Economic Performance’.

For fewer crises, see this paper's DCSI model; for less extreme crises, Acemoglu et al., ‘Institutional Causes’.

For example, Pastor and Sung, ‘Private Investment and Democracy’; Feng, ‘Democracy, Political Stability, and Economic Growth’; Heinsz, ‘Political Institutions and Policy Volatility’; Chong and Calderon, ‘Causality and Feedback’.

As in Stein et al., ‘The Politics of Policies’.

For example, Acemoglu, Johnson, and Robinson, ‘Colonial Origins of Comparative Development’; Acemoglu et al., ‘Institutional Causes, Macroeconomic Symptoms’.

Acemoglu et al., ‘Institutional Causes, Macroeconomic Symptoms’, 61–7.

RL theorists are impressed by the example of a few soft authoritarian, yet apparently law-governed, regimes such as contemporary Singapore. We suggest that informal yet increasingly institutionalized processes of interest-aggregation and proto-contestation, including regular consultation with the business community, may make Singapore about as ‘democratic’ as early nineteenth-century England.

Acemoglu et al., ‘Institutional Causes, Macroeconomic Symptoms’, 68.

Gleditsch and Ward, ‘Double Take’.

Tsebelis, ‘Decision Making in Political Systems’.

Heinsz, ‘Political Institutions and Policy Volatility’; Cox and McCubbins, ‘Institutional Determinants of Economic Policy’; Mobarak, ‘Democracy, Volatility, and Economic Development’, 351.

Cf. Almeida and Ferreira, ‘Democracy and the Variability of Economic Performance’.

Chandra and Rudra, ‘Regime Type and Economic Performance’; Lindblom, Intelligence of Democracy; cf. Rodrik, ‘Participatory Politics’.

Przeworski, Democracy and the Market, 10.

Alesina and Rodrik, ‘Distributive Politics and Economic Growth’; Persson and Tabellini, ‘Is Inequality Harmful’.

Baer, ‘Social Aspects’; Hirschman, ‘Social and Political Matrix of Inflation’; Dornbusch and Edwards, Macroeconomics of Populism.

Quinn and Wooley, ‘Democracy and National Economic Performance’; cf. Almeida and Ferreira, ’Democracy and the Variability of Performance’, 225–8.

Weyland, Politics of Market Reform.

Bueno de Mesquita et al., ‘Political Institutions, Political Survival’; Morrow et al., ‘Retesting Selectorate Theory’.

Bueno de Mesquita et al., ‘Political Institutions, Political Survival’, 60–1, 70, 61.

Ibid., 64; Morrow et al., ‘Retesting Selectorate Theory’, 393.

Morrow et al., ‘Retesting Selectorate Theory’, 395.

Ibid., 395; Marshall and Jaggers, Polity IV Project.

Moon et al., ‘Voting Counts’; Coppedge, Alvarez, and Maldonado, ‘Two Persistent Dimensions’.

Dahl, Polyarchy; Dahl, Democracy and Its Critics; cf. Vanhanen, ‘A New Dataset for Measuring Democracy’.

Przeworski et al., Democracy and Development, 176–8; Quinn and Wooley, ‘Democracy and National Economic Performance’.

Rueschemeyer, Huber Stephens, and Stephens, Capitalist Development and Democracy.

Cf. Halac and Schmukler, ‘Distributional Effects of Crises’.

Popkin, The Rational Peasant.

Cf. Bueno de Mesquita et al., ‘Political Institutions, Political Survival’; Morrow et al., ‘Retesting Selectorate Theory’.

Thus our ‘politically relevant actors’ correspond to the ‘winning coalition’ of Bueno de Mesquita et al., not to the ‘selectorate’, the members of which vote and participate symbolically, but lack substantive power either to choose or depose/impeach incumbents.

Eichengreen, Globalizing Capital; Simmons, Who Adjusts?

Drèze and Sen, Hunger and Public Action.

Przeworski et al., Democracy and Development.

Coppedge, Alvarez, and Maldonado, ‘Two Persistent Dimensions’.

The Polity dataset is the political science discipline standard for measuring democracy, especially over long time periods. See Marshall and Jaggers, Polity IV Project. However, it recently has been the object of several critical evaluations. These include Gleditsch and Ward, ‘Double Take’; Munck and Verkuilen, ‘Conceptualizing and Measuring Democracy’; and Treier and Jackman, ‘Democracy as a Latent Variable’.

Moon, et al., ‘Voting Counts’; Coppedge, Alvarez, and Maldonado, ‘Two Persistent Dimensions’.

Franklin, ‘Electoral Participation’.

Verba and Nie, Participation in America; Hill and Leighly, ‘The Policy Consequences of Class Bias’; and Griffin and Newman, ‘Are Voters Better Represented?’ demonstrate this phenomenon for the US.

Lijphart, ‘Unequal Participation’; Chong and Oliveira, ‘Does Compulsory Voting Help Equalize Incomes?’

Pacek and Radcliffe, ‘Turnout and the Vote’; Hicks and Swank, ‘Politics, Institutions, and Welfare Spending’.

Paxton et al., ‘A Half-Century of Suffrage’; Vanhanen, ‘A New Dataset for Measuring Democracy’.

For the 1960–98 period, the authors code only 3.9% of the country-years as giving voting rights to more than 0%, but less than 90%, of the adult population.

Given that Vanhanen's denominator is not the voting age population but the total population, the 50% threshold is very demanding. Only 19.7% of country-years lie above this level.

Moon, et al., ‘Voting Counts’.

Coppedge, Alvarez, and Maldonado, ‘Two Persistent Dimensions’.

Ibid., 632.

Acemoglu et al., ‘Institutional Causes, Macroeconomic Symptoms’.

The minimum number of countries with data for the main dependent variable for a given year is 70 (1961) and the maximum 148 (1993–94). The moderate amount of missing data is concentrated in poor and Communist countries, and in earlier years.

Macroeconomic data are from World Bank's World Development Indicators, various years.

Political regimes change only infrequently, so most annual Polity scores are the same as or very similar to the average of the previous five years (Pearson's r for these two variables is 0.96).

The expectation here is also curvilinear because of diminishing marginal returns to investment.

Fearon, ‘Ethnic and Cultural Diversity’; cf. Easterly and Levine, ‘Africa's Growth Tragedy’.

Beck, Katz, and Tucker, ‘Taking Time Seriously’; Beck, ‘Time-Series-Cross-Section Data’.

Wawro, ‘A Panel Analysis of Campaign Contributions’, 567.

Zorn, ‘Generalized Estimating Equation Models’.

Hsiao, Analysis of Panel Data.

Zorn, ‘Generalized Estimating Equation Models’.

Fitzmaurice, Laird, and Ware, Applied Longitudinal Analysis. On one criterion GEE models are inferior to random and fixed effects models: their sensitivity to missing data. With data missing at random, GEE produce biased coefficients, but likelihood-based methods such GLMMs do not, assuming that the model and the random effects are correctly specified (ibid., 385). This qualification makes the missing data advantage of these models a tenuous foundation to prefer them over GEE models.

In all GEE models we used robust standard errors. Stata's ‘robust’ option for the xtgee command automatically clusters standard errors by the variable identifying cross sectional units (countries in our case). Clustered standard errors are correct even if some serial correlation remains after that modeled in the GEÉs correlation structure. The corresponding Stata commands combine ‘xtgee’ with the options ‘fam(bin)’, ‘link(logit)’, ‘corr(ar1)’, ‘corr(sta6)’, or ‘corr(ind)’, and ‘robust’. To avoid losing hundreds of observations due to the gaps in our database, we used the option ‘force’. In the stationary(6) specification several countries are dropped because they have fewer than 7 observations. We tried running the model with an unstructured correlation matrix, but the large number of parameters to estimate made convergence impossible. The autoregressive structure makes the within-unit correlations an exponential function of the lag length (cf. Zorn, ‘Generalized Estimating Equation Models’). Correlations are constrained to be constant across observations for each lag length, and to equal ρ¹ for a lag of one year, ρ² for two years, and so forth. The stationary structure also makes correlations across observations constant for a given lag, but estimates ρ for each of them from the data. Since this implies calculating more parameters, it is common to specify a maximum number of lags. We made this number 6 as our exploration of the data indicated that beyond six years the correlations become essentially random noise around zero. Both correlations structures are appropriate to model processes that, like crises, are expected to have positive but declining temporal dependence.

Standardized coefficients allow for the comparison of variables with different scales and standard deviations. Democracy and Democracy(PEPS) share a -10 to 10 scale, but the former has a somewhat smaller standard deviation. X-standardized coefficients are calculated by multiplying the unstandardized GEE logit coefficient of X_i by the standard deviation of X_i (that is, the SD in the effective sample used in a given model).

Greene, ‘Econometric Analysis’, 154.

Figure 1 was constructed using the coefficients of model 2 in Table 3.

All these models except random effects are run with standard errors clustered by country.

A total of 41 and 48 countries (in table 4 and 5, respectively) are dropped in the fixed effects models due to no variance on the dependent variable, reducing N from 3,285 to 2,573 and from 2,515 to 1,790, respectively. These countries include most of the advanced industrial democracies, but also many less developed nations such as Egypt, Pakistan, and Sri Lanka. Remarkably, Turkmenistan was also dropped due to all positive outcomes (it experienced a recession every single year from its separation from the USSR until 1998).

Przeworski, Democracy and the Market, 136–92.

Bernhard and Leblang, Democratic Processes and Financial Markets, 228–31; Keefer and Stasavage, ‘Checks and Balances’.

Arbetman and Kugler, Political Capacity and Economic Behavior; Evans, ‘The State as Problem and Solution’; Kohli, State-Directed Development.

Cf. Sen, Democracy as Freedom.

Cf. Burke, Reflections on the Revolution.

See O'Donnell, ‘Delegative Democracy’; Zakaria, ‘Illiberal Democracy’; Diamond and Morlino, ‘Assessing the Quality of Democracy’.

Dahl, Polyarchy.

Cf. Armijo, ‘Mass Democracy’; Slater, ‘Can Leviathan Be Democratic’.