The state of democracy in Sub-Saharan Africa

Africa experienced a wave of democratization over the past 20 years and this increase in democracy, we find, positively and significantly affects income per capita. Our dynamic panel data results suggest that countries only slowly converge to their long-run income values as predicted by current democracy levels, however. African countries may therefore be currently too democratic relative to their income levels. In keeping with this possibility, a significant number of countries are experiencing political ‘back sliding’: elections are won by the use of illicit tactics, term limits on political leaders have been overturned and there have been unconstitutional seizures of power.


Introduction
In this paper we examine the state of democracy in Sub-Saharan Africa and its impact on Africa's economic performance.
Following 1989 the region experienced a wave of democratization: most countries held elections and legalized multi-party systems and many introduced term limits for their leaders.
Subsequently, Africa's economies grew at a rate not witnessed since the time of independence. We offer evidence of a causal relationship between democratization and economic performance. We also offer evidence that the democratic impulse may have weakened and economic growth now arises from economic forces external to Africa rather than from political forces within.
Section 2 provides a brief overview of the literature on the relationship between income and democracy and the place of Africa within it. In Section 3 we discuss our methods and data and, in Section 4, our statistical results. We find that for Sub-Saharan Africa democracy 'Granger' causes income. We then quantify the long-as well as the short-run relationship between democracy and income. Our error-correction mechanism predicts a slow adjustment to the long-run equilibrium, which, we suggest, implies that recent levels of democracy may be too high for prevailing levels of income. In support of this conjecture, we present recent evidence on political 'back sliding. Section 5 concludes.

Background
The analysis of democracy in Africa can usefully be placed within the larger literature on income and democracy. Seymour Martin Lipset reported a strong and positive correlation between income per capita and democracy in a global cross section of nations (Lipset 1959).
He suggested that economic development causes a series of profound social changes that result in democracy. Doing so, he not only lay the foundations of modernization theory in comparative politics but also defined a major portion of the contemporary agenda in political economy, with its focus on the relationship between political institutions and economic development. 1 The new institutionalists (e.g. North and Thomas 1973;North 1981;North 1990) take a different view: Unlike Lipset they suggest that good institutions cause development (see also Barro, 1996). The impact of the new institutionalism extends beyond academic circles. By way of illustration, consider the work by Burnside and Dollar (2000), who suggested that aid is only growth enhancing in environments blessed with good institutions. In response to institutionalist arguments, donors began to offer aid selectively, i.e. to countries with good governance. 2 As a result, in Africa, a number of countries received less aid.
Dissenting from both schools, recent contributors suggest that there is no relationship between income and democracy. Przeworski et al (2000) failed to find a significant relationship between the level of income per capita and the likelihood of transition to democracy. While Boix and Stokes (2003) and Epstein et al (2006) have challenged correlation between democracy and income, they claim that there is no evidence for a causal link. Their panel regressions show that higher incomes do not lead to higher levels of democracy. They interpret their findings as evidence "that political and economic political development paths are interwoven" (AJRY: 836) and that at "certain critical junctures" (AJRY: 813) societies embarked on divergent political-economic development paths. A result of these divergent processes is that some countries end up democratic and rich while others remain autocratic and poor. While this appears plausible, their interpretation may rest on weak foundations: Gundlach and Paldam (2009) argue that AJRY find no relationship between democracy and income due to the statistical methods they apply.
The study of Africa has both much to learn and much to contribute to these debates. From them, it can learn how best to measure the relationship between political change and economic performance: an issue that we confron in the section that follows. In addition, it can help to adjudicate the debate over Lipset's conjecture, evenexploring the state of democracy in Sub-Saharan Africa.

Methods and Data
Past work has either examined a cross-section of countries (Lipset, 1959) or panels containing a large number of countries ('large n') but a small number of years ('small t') (e.g. Barro 1996Barro , 1999AJRY). We focus on the relationship between income and democracy, 3 measures of which exist for most countries and extend back to 1960. By comparison with the methods employed by previous researchers, our estimator makes more efficient use of time series data. Our ability to make greater use of the temporal dimension of the data enables us (1) to investigate the direction of causality and (2) to examine the long and short-run relation between income and democracy in a panel error correction model.

Data
We use the Penn World Tables' (PWT 6.3) chain weighted real GDP per capita series and the Polity IV democracy index which distributes over a range spanning the interval between perfect autocracies (score of -10) and perfect democracies (score of 10). Our sample includes 105 countries, 42 of which are in Sub-Saharan Africa 4 . Figure 1 shows that on average incomes and polity scores have risen over time. While incomes have grown relatively smoothly, in the late 1980s, the polity index jumped discontinuously from -0.4 in 1989 to 1.9 in 1992. As seen in Figure 2, there are important regional differences in the movement toward democracy. Latin America democratized prior to the fall of Communism. Africa and the Middle East both democratized after 1990. The polity scores then diverged, with those in Sub-Saharan improving more rapidly.

Methods: Granger Causality Tests and Pooled Mean Group (PMG) Estimation
We begin by entering into the first of the debates outlined above and ask: Does income cause democracy or does the causal relationship run the other way 'round? We use 'Granger causality' tests to tackle this question 5 . The idea of 'Granger causality' is that if income causes democracy then (1) income should help to predict democracy and (2) democracy should not help to predict income. To test the null hypothesis that 'income does not cause 4 The sample size is limited by the method we employ. We use Pooled Mean Group (PMG) estimation which computes coefficients for each country separately. Thus we can only include countries with sufficiently long time series and cannot include countries with no time variation in the dependent variable. 5 For a textbook description of this method see for example Wooldridge (2009). democracy' we regress democracy, d, against lagged values of democracy and lagged values of income, y ('the unrestricted regression'): . We then regress democracy only against lagged values of democracy ('the restricted regression'): A simple F test can then be used to determine whether the lagged values of income contribute significantly to the explanatory power of the 'unrestricted model'. If they do, we reject the null hypothesis and conclude that income 'Granger' causes democracy.
We also test the null hypothesis that 'democracy causes income': versus .
As Table 1 shows, for the global and non-Sub-Saharan Africa samples, Granger tests indicate that for the global sample causality between income and democracy runs in both directions.
But for the Sub-Saharan Africa portion of the sample they indicate that democracy 'Granger' causes income. Our findings thus indicate that while income and democracy are positively related in the global sample, the relationship is not causal, but that in Sub-Saharan Africa, democratization has produced higher incomes.
--- Table 1 about here ---Having established the possibility of causality, we estimate the short and long term relationship between income and democracy, focusing primarily on Africa. We could follow AJRY and use either pooled OLS or fixed effects estimation. However, we decide to use an augmented version of the Pooled Mean Group (PMG) estimator of Pesaran et al (1999) for a number of reasons. The end of the Cold War resulted in an exogenous worldwide wave of democratization and we want to analyze the dynamic response of incomes and democracy levels across countries. 6 This analytic focus requires the use of a dynamic panel estimation technique that allows us to make full use of the available time series data. Furthermore, OLS as well as fixed effects estimation assume that the parameters are homogenous across the panel, i.e. all of the countries respond to changes in the same way. In contrast, the PMG estimator not only allows us to account for country and year effects but also for parameter heterogeneity across panel members. We now discuss our choice of estimator in more detail.
In their critique of AJRY, Gundlach and Paldam (2009) suggest that the inclusion of time and country fixed effects purged useful information in panel data estimation, thereby predisposing them to fail in their search for a relationship between income and democracy. Their argument highlights an important methodological dilemma: Including country specific fixed effects eliminates useful informative variation from the data; but excluding them introduces omitted variable bias. Employing an augmented version of the PMG estimator of Pesaran et al (1999), we confrontand surmountthis dilemma. While taking into account country and year effects, we relax the assumption of cross-sectional parameter homogeneity. Even while controlling for omitted variables, we thereby extract information from sources of variation that, with their methods, AJRY had perforce to ignore.
The PMG estimator allows intercepts, slope coefficients and error variances to vary across countries. More specifically, it allows the short-run coefficients to vary across countries, while restricting long-run relationships to be homogeneous. In the context of this research, the estimator "assumes" that in the short runor while adjusting to a common long-run equilibrium -each country's political institutions respond differently to income shocks.
Because it allows for heterogeneous intercepts, the PMG estimator can incorporate countryspecific fixed effects. But because it estimates the model for each country separately, it cannot allow the inclusion of year fixed effects. To correct for potential cross-section dependence in the estimated errors, weas do Binder and Offermanns (2007) Crucially, the error term it  is identically and independently distributed across i and t even in the presence of common time effects. Country intercepts --unobserved country heterogeneity are captured by the term i  .
The second part of equation (1) includes the lagged changes of income and democracy; the coefficients represent the short-run adjustment terms and are assumed to vary across countries. We do not report the short-run coefficients below. The first part of equation (1) captures the common long-run relationship between income and democracy. The slope coefficients -- , , and  --measure the long-run response of income to democracy, world income and world democracy.  is the error correction coefficient and indicates the speed of adjustment If the system is dynamically stable and converges to a long-run equilibrium, then this coefficient will be negative and less than one in absolute value. We report these long-run coefficients below.
Starting with an initial estimate of the long-run parameters, the PMG estimator calculates estimates of error-correction and other short-run coefficients (including country-specific intercepts and error variances) as the averages of the estimated parameters for each crosssection. It then employs these average estimates to update its estimates of the long-run parameters, repeating the process until convergence is achieved.

Results
The results of our analysis confirm the existence of a positive relationship between the level of democracy and income in Sub-Saharan Africa (Table 3). 8 They suggest that a one-unit increase in democracy leads to a 1.5% increase in per capita income. Given that the average

Why Africa?
With the data at hand, we are not in a position to explain why political reform led to economic growth in the African but not in other portions of the global sample. We are in a position to offer hypotheses, however. Our favourite is based on the high degree of urban bias that was exhibited by Africa's authoritarian regimes and the realization that democratization led to the enfranchisement of a largely rural electorate.
As summarized in Ndulu et al (2008), the economic policies of many African regimes were characterized (inter alia) by:  Tariff policies that protected domestic manufacturing (but not agriculture).
 Industrial regulations that conferred market power on the purchasers of agricultural products rather than on the producers of manufactured goods.
 Over-valuation of their domestic currencies exchange rate.
Given that manufacturing received offsetting protection from foreign products, the last of these measures further tilted relative prices in favour of the urban sector. Taken together, the policies were therefore biased against agriculturethe largest single industry in most of Africa's economies in Africa. One result was slow growth. The estimates reported in Ndulu et al (2008) suggest that governments that adopted this mixture of policies lowered their country's rate of growth by nearly two percentage points per annum 1960-2000.
That political change led to policy change is suggested by data reported in Figure  intervened in markets in ways less likely to shift relative prices against farmers (as indicated by their Relative Rates of Assistance) 11 ; and that they spend more on agricultural research, secured higher levels of educational attainment, and paved a larger percentage of their roads 12 . Calculating the means, we apply one sided t-tests to the differences and find each to be significant and in the expected direction. Governments in competitive political systems acted in ways that lower the costs, increase the earnings, and strengthen the incentives for farmers. Given the importance of agriculture in Africa's economies, it is not surprising that economic growth followed the choice of such policies.
To close the argument, we need but note that outside of Africa the primary locus of political reform took place lay in Eastern Europe and the former Soviet Union, where the structure of the economies differed greatly from those in Africa. Particularly when the transition between 10 For a multivariate exploration, see Bates and Block (2010). A competitive political system is defined as one in which the head of state was voted into office in an election in which an organized opposition party can and did run a rival candidate who received at least 25% of the vote. 11 The variable provides a comparison of the impact of government policies on the relative prices of agricultural and non-agricultural commodities. Lower values imply greater urban bias (Anderson, 2010). 12 Most school aged children live in the rural areas and higher transport costs result in lower farm-gate prices. socialist and market-based economies was impeded by the actions of large firms and their allies in the bureaucracy (Hellman 1998), it remained politically attractive to continue to advocate policies that favored consumers over producers in the post-socialist economies.
Beneath this argument lie assumptions about "the African voter," and, in particular, that rural voters are willing and able to respond to policy positions rather than, say, communal appeals.
A rapidly growing number of studies suggest that while individual voters might prefer candidates who come from their ethnic group, many fail to find the opportunity to vote in accord with that preference. 13 Researchers also find that evaluations of the economy play a major role in voting decisions, Given these findings, politicians whose fates rest in the hands of a rural electorate might well be loath to advocate policies that favor urban consumers over rural producers. 14

Reasons for caution
The results also suggest several reasons for concern. When the sample is disaggregated by time period (Table 3), the coefficient relating democracy to income appears to decline, suggesting that the relationship between the level of democracy and income has eroded. 15 Troubling too is the magnitude of the error correction coefficient (Table 2): -0.122, which implies that the response slow, taking nearly a decade to accomplish. 16 13 It is not possible for parties to generate the quantity and mixture of candidates to satisfy such preferences. 14 See Mattes and Piombo (1999), Posner and Simon (2002), Gibson and Long (2012) and Hoffman and Long (2012). 15 Indeed, the data in Table 3 suggest that Africa's current economic growth appears now to be propelled by growth abroadas in China and India, for example-rather than by political reform at home. 16 These results are robust to the inclusion of foreign aid as a percent of GDP as an additional covariate. Inspection of the estimates of the individual country error correction coefficient suggests that Angola, Burkina Faso, Gabon, Gambia, Ghana, Guinea Bissau, Liberia, Malawi, Mali, Mauritania, Mozambique, Nigeria, Rwanda and Tanzania exhibit higher than average values, suggesting that income adjusts more quickly to its long-run value, given their current level of democracy. Rwanda, Mauritania, Ghana and Gambia have error correction coefficients three or more times higher than the average. The reasons for this dispersion continue to intrigue -and to elude -us.
Taken togetherand possibly over-interpretingthese findings may be suggesting that political reform left Africa "too democratic," given the level of income. They may imply that the continent has begun to regress to its expected value of democracy, not by growing economically but rather by becoming less democratic. While confirming the political origins of economic growth in Africa, our results thus also suggest their fragile nature.
Data from other sources deepen this concern. Consider Figures 2, 4 and 5: Surging upwards in the late 20th century, the Polity index for Africa's governments continued its ascent in the 21st, albeit at a lesser rate. But as the first decade of the new century ended, political progress ended as well, stalling out at an average country score of 2 in a scale that runs up to 10. Qualitative accounts, moreover, confirm that Africa's governments, intent upon slipping the bonds of electoral accountability, increasingly abuse political rights and civil liberties. As

Conclusion
Using a panel error correction estimation method we have found, contra Lipset (1959), that in Africa democracy elicits economic growth, rather than the other way 'round, lending support to a "new institutionist" interpretation of Lipset's hypothesis. The last-century wave of democratization appears to have resulted in increased incomes across the region. The enfranchisement of Africa's rural majorities appears to have focused the minds and altered the policy preferences of its governments, resulting in more favorable conditions for agriculture, the largest single industry in most African economies, and higher rates of growth in Africa's agrarian economies.
Focusing on Africa itself, we have learned that the re-introduction of competitive electoral systems appears to have enhanced the level of income. However, the data appear also to suggest that the political origins of Africa's growth remain fragile. And, indeed, in a large number of countries these reforms appear embattled: elections are won by the use of illicit tactics, term limits have been challenged and, in some instances, abolished; and unconstitutional leadership turnovers remain distressingly commonplace. Given the longterm relationship between income and democracy, recent democracy levels may in Africa have been too high to be sustainable; and international rather than internal forcessuch as the level of global income --may determine its economic future.  1962 1965 1968 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 Polity Score

Polity
Income per capita Does the constitution provide a two term limit for the presidency?
Has that term limit been reached?
Was there an attempt to amend the constitution?
Did it succeed?
Eq. Guinea Adapted and extended to March 2012 from Posner and Young (2007). 1 Seychelles has a three term limit. Note: In testing whether democracy Granger causes income, income is regressed on lags of income and democracy, and the reported F-stat is a Wald-type test of the joint significance of all estimated coefficients on such lags. We also report the probability of rejecting the null hypothesis. Notes: All equations include a constant country-specific term. Numbers reported in parentheses are standard errors. Numbers reported in brackets are p-values.***, **, and * indicate significance respectively at the 1, 5, and 10 percent levels. We use the Schwartz Bayesian optimal lag selection Criterion subject to a maximum lag of three. World democracy and world output are respectively the cross-sectional averages of democracy and output, which we take as proxies of the common unobserved global shocks. Notes: All equations include a constant country-specific term. Numbers reported in parentheses are standard errors. Numbers reported in brackets are p-values.***, **, and * indicate significance respectively at the 1, 5, and 10 percent levels. For brevity we only report PMG results. The small time series dimension allowed us to impose a common lag of one on income and democracy instead of suing optimal lag selection criteria. World democracy and world output are respectively the crosssectional averages of democracy and output, which we take as proxies of the common unobserved global shocks. The cross-sectional dimension varies for each column since over each time period the countries who happen to have constant polity scores drop out.

Democracy
Polity IV index ranging from -10 for perfect autocracies to +10 for perfect democracies.