Revisiting African Agriculture: Institutional Change and Productivity Growth

Full Text

 Revisiting African Agriculture: Institutional Change and Productivity Growth

 

 

 

 

 

Abstract

Africa is largely agrarian and the performance of its agriculture sector shapes the performance of its economies.  Building on a recent analysis of total factor productivity growth in African agriculture, we explore the politics underlying the economics of this sector.  In so doing, we explore the relationship between political institutions and economic development.  The introduction of competitive presidential elections, we find, is significantly related to the resurgence of productivity growth in African agriculture.  The emergence of electoral competition appears to have altered political incentives, resulting in both sectoral and macroeconomic policy reforms that strengthened productive incentives in farming.

 

 


1.   Introduction

 

In the later decades of the 20th Century, political institutions in Africa changed.  Prior to the late 1980s, open competition for national office was rare: politicians became heads of state either by launching military coups or by consolidating their political backing within the ruling party.  Subsequently, most heads of state were instead chosen in elections contested by rival parties that competed to capture political support from a majority of the national electorate. 

 

Figure 1 documents the nature and magnitude of these changes.  Classifying political systems along a 7-point scale that captures the level of electoral competition (see below), the figure depicts the striking shift towards competitive politics.  In the 1970s, the mean lay below 3; by the 21st century, it lay above 6. 

 

In general, institutions change slowly and by few degrees and this lack of variation renders it difficult to study their impact (Persson and Tabellini 2008).  The speed and scope of Africa’s political transformation thus renders the continent a potential source of evidence concerning the relationship between political institutions and the performance of economies. 

 

On average, one third of Africa’s people work in farming and agriculture remains the single largest industry in most of Africa’s economies.  The decline of the rural sector in the 1970s foreshadowed Africa’s economic collapse (World Bank 1981); agriculture’s current revival lends impetus to the continent’s economic recovery.  It is our claim that changes in Africa’s political institutions have shaped this economic trajectory. 

 

The Literature

 

More broadly, this study engages the so-called “Lipset hypothesis” (Lipset 1959). Observing a relationship between economic development and democracy in a global cross section of nations, Lipset (1959) posited that the former caused the latter.  Using panel data and applying a Markov model, Przeworski et al. (Przeworski, Alvarez et al. 2000) challenged Lipset’s argument, arguing that richer countries were no more likely to become democratic; rather, they were more stable and, if democratic, less likely than poor countries to turn authoritarian (but see Boix and Stokes 2003).  Applying a dynamic OLS estimator with fixed effects on panels that ran far deeper than those employed by either Lipset (1959) or Prezeworski et al. (2000), Acemoglu, Johnson et al. (2008) too found no significant relationship between democracy and development.  Subsequently, Persson and Tabellini (2008) applied matching techniques to a contemporary cross-section of nations and failed to detect significant differences between levels of democracy and rates of economic development.  Particularly among authoritarian countries, they report, the “within” variance in growth rates exceeds the “between,” thus accounting for their finding.

 

Undeterred by this legacy of negative findings, we return to Lipset’s argument, but in a manner that differs in key respects from those who have gone before.  Rather than using global samples, we restrict ourselves to a regional sample and make use of data from Africa.  In contrast to other regions of the world, in Africa, institutions are unstable.  While disadvantageous in many respects, this instability offers advantages to those seeking to explore whether institutional differences bear upon differences in development.  In addition, with the exception of Przeworski et al. (2000), investigators have tended to employ the Polity measure of democracy; and all have employed the gross domestic product per capita as a measure of economic development.  The Polity score is derived from data on several institutions, however; in addition, it is subjective, recording such attributes as the quality of political competition and the degree of political openness.  The Polity score thus lacks sufficient “micro”- level precision to give rise to strong expectations as to how a change in political institutions would alter the behavior of policy makers, the policies they might choose, and thus the impact of institutions upon the performance of the economy.

 

In response, while grounding our analysis on African data, we make use of alternative measures of political institutions and economic development.  Instead of the Polity measure, we employ a measure first developed for the analysis of political transitions in Africa and subsequently collected for a global set of nations by the World Bank (Beck, Clarke et al. 2001).  Focusing on the manner in which chief executives secure political office, the measure can be thought of as a “Schumpeterian” (Schumpeter 1950) index of political competition: the higher the number, the greater the degree of party competition.[1]  Because the measure does not take into account the degree to which this competition takes place in an environment in which there are safeguards to civil and political rights or to which elections, though competitive, are free and fair, it offers little insight into the quality of democratic government, however (Dahl 1971).

 

Despite this deficiency, when applied to the political systems of Africa, the measure suggests important insights into the nature of the political incentives confronting those aspiring to national office and the nature of the political strategies that they are likely to adopt.  Africa is poor and its urban population and industrial centers are small, while its rural population and agrarian sectors are large.  In the absence of party competition – i.e. when the value of our measure is low – we should therefore expect economic interests to seek to influence public policies through lobbying by interest groups.  Insofar as this is the case, we expect economic policies to be “urban biased.”  But when there is a change to party competition – i.e. when the value of our measure becomes high – electoral majorities acquire political weight and those seeking office will therefore begin to advocate measures designed to attract rural voters.  They should therefore begin to support policies that favor agriculture.  The change from interest group politics to electoral politics should therefore induce a policy change; it should induce a shift in relative prices in favor of agriculture, thereby rendering agriculture more profitable and strengthening the incentives for farmers to make more productive use of factors of production.

 

Moving from “macro-“ level indices – such as the Polity score and GDP per capita -- to “micro-” level measures – of political institutions and incentives – thus suggests a line of argument linking political institutions and economic performance.

 

Note a key assumption underlying this argument: that “the African voter” engages in performance voting and that Africa’s politicians therefore need fear political punishment should they fail to adopt policies desired by a majority of the electorate.  One counter argument would see ethnic identity trumping economic self-interest in the minds of voters; a second would see distributive politics –clientelism and vote buying – as dominating policy preferences.  Taken together, the two would suggest that politicians in Africa need not champion public policies when competing for electoral majorities and that the expectations that underlie our argument are therefore unjustified. 

 

Recent research suggests that while there is an element of truth to both arguments, politicians nonetheless remain accountable for their policy choices.  In a recent paper, Philip Keefer (2010) marshals data from 16 Afribarometer surveys and finds little evidence that ethnic groups systematically align with political parties.  In a study of the 2008 election in Ghana, Ferree et al. (2009) find little evidence of bloc voting by ethnic groups; and while politicians did make ethnic appeals, the authors find that voters focused on the policies and the competence of the political parties rather than on the ethnic identity of the candidates.  In a study of the 1999 elections in South Africa, Mattes and Piombo (1999) present evidence suggesting that voter decisions result from a comparative evaluation of the incumbent and opposition party, in which the incumbent is evaluated retrospectively and the opposition prospectively.  In making these evaluations, they argue, voters focus on the “direction of the economy” and the “direction of the country” (p. 110).  Race, which is akin to ethnicity in South African context, may affect the manner in which economic conditions are evaluated, they find, but not their importance to the voter.[2]  In their study of the Zambian election of 2002, Posner and Simon (2002) find that while ethnicity affects voter preference, so too does the voters’ satisfaction with the condition of the economy; those who “expressed dissatisfaction … were 10 to 15 percentage points less likely to vote for the incumbent” (p. 319) – i.e. to hold the government politically accountable.  The magnitude of this effect was greater than that associated with ethnic differences.[3] 

 

As argued by Vicente and Wantchekon (2009), when political parties fail to act as agents of ethnic interests, then political organizers must find other means for delivering votes.  Incumbents use government agencies and patronage to mobilize votes; being unable to do so, challengers seek to buy them.  Insofar as these strategies work – and Vicente and Wantchekon (2009) argue that they do – they undercut incentives for politicians to advance policies capable of benefitting electoral majorities.  As noted by Stokes (2005), however, vote buying – through either cash or clientelism – requires perfect information; resources would be wasted if targeted on people with strong partisan attachments and politicians must be able to verify that a vote, once paid for, stays bought.  Judging by Stoke’s (2005) results, even in Argentina, where the Peronists have long been held to have mastered the art of clientelist politics, perfect information is difficult to achieve.  Parties appear to garner less than 10% of their votes in this fashion.

 

While clientelism, like ethnicity, may be productive electorally, its overall impact may therefore not be large.  Expected policy benefits– i.e. the content of policies and the competence with which they are implemented – appear to play the dominant role in voter decisions.  The assumption we make about “the African voter” thus appears to be justified, judging by recent research.

 

Our paper relates to the work of Stasavage (2005) and Kudamatsu (2007).  Working with data from 44 African countries, 1980-1996, Stasavage (2005) found that governments chosen in elections openly contested by rival political parties spend more on primary education.  Political reform led to higher levels and more geographically dispersed service delivery, he contends.  Working with household-level data from 28 African countries, Kudamatsu (2007) found lower levels of infant and neo-natal mortality for children born following the introduction of competitive elections.  As did Stasavage (2005), he attributes the change to improvements in service delivery, as politicians respond to the need to secure votes from an enfranchised citizenry.  By focusing not only on policy change but also upon its impact, we extend the agenda set by Stasavage (2005) and Kudamatsu (2007).  We do so by advancing evidence that institutional change appears to have promoted the growth of total factor productivity in farming.

 

Our research also contributes to the literature on democracy and development (Persson and Tabellini 2008).  In contrast with Barro (1996), when estimating the relationship between democracy and growth, we allow for the possibility of reciprocal causation (Lipset 1959).  In doing so, we dissent from Glaeser, La Porta et al. (2004): we find that democracy is related to growth, not because growth induces a demand for political reform, as they contend, but rather because, in the political and economic context which we address, democratization induced changes in policy, thereby altering economic incentives.  Like Giavazzi and Tabelinni (2005), we acknowledge that “the devil is in the details;” because our investigation is narrowly targeted, we are better able than they to isolate and trace the channels linking institutional change to variation in economic performance.

 

Section 2 presents our argument.   Section 3 provides political and economic background, portions of which provide justification for our estimation strategy.  Section 4 describes our data, while Section 5 presents our estimates of the relationship between institutional change, policy reform, and the economic performance of agriculture.  Section 6 concludes.

 

2. Engel’s Law, Structural Transformation, and the Logic of Political Representation

The relationship between political reform and economic change in Africa can be derived from well-established insights into the consumption behavior of poor persons and the structure of their economies on the one hand and from the logic of collective action and party competition on the other.

 

Engel’s law holds that as income rises, the proportion of income spent on food declines; the income elasticity of food consumption is less than unity.  From this micro-level regularity a macro-level implication follows: that economic development implies structural change (Kuznets 1966; Chenery and Taylor 1968).  When people are poor, a large percentage of their total expenditure will be devoted to food; absent foreign trade and significant economies of scale in farming, the rural sector will therefore be large.  But when people earn higher incomes, the percentage spent on food will be less and the rural sector will then comprise a smaller portion of the economy.   The process of economic development is thus marked by the marginalization of agriculture.

 

Given commonly observed characteristics of farms and firms, poor countries therefore exhibit a characteristic political-economic geography.  The majority of the population works in farming; it lies widely scattered, each member producing but an infinitesimal percentage of the total output of agricultural goods.  A small portion of the population – often less than 10% -- works in manufacturing and service provision and dwells in towns.  Because economies of scale exist in manufacturing, a significant percentage of the urban dwellers derive their incomes from a small number of employers (Little, Scitovsky et al. 1970; Little 1982).  While those who farm are thus dispersed, economically and geographically, those who earn their incomes in other portions of the economy are not.  Spatially, they are concentrated in a few settlements and economically they often labor in enterprises sufficiently large to dominate their markets. 

 

The political implications are immediate and ironic: In countries with large agricultural populations, farmers form a weak political lobby.  Being “small,” farmers in poor countries rationally refrain from expending effort in attempts to influence agricultural prices; not so urban interests, which stand “large” in their markets.  Being widely scattered, farmers face high costs of organizing; concentrated in towns, urban interests find it less expensive to do so.  Urban interests therefore hold a relative advantage as lobbyists in less developed economies.  In so far as government policy is influenced by organized groups, in countries with large agricultural sectors, it tends to be adverse toward the interests of farmers (Olson 1971, 1985).

 

But now, consider: What if the process of representation should follow the logic of electoral competition rather than collective action?  Then the very factors – size and dispersal – that render farmers weak lobbyists would render them powerful (Varshney 1995).  Where representation is achieved through electoral channels and where rural dwellers constitute a large segment of the voting population, then politicians have an incentive to bear the costs of political organization and to cater to the interests of farmers.  The search for political majorities should lead them to resist the political pressures emanating from urban consumers and to champion policies that cater to the interests of the rural majority.

 

The mean income in Africa is less than $1,000 per annum (constant $US 2,000).  Consistent with Engel’s law, agriculture is the largest single industry: on average, one third of its nations’ workers derive their incomes from farming and three quarters of its population resides in the rural areas.  On the basis of the reasoning above, then, we would expect that the implications of this argument would follow and that changes in Africa’s institutions would generate concomitant changes in the behavior of its governments.


3. Background

 

By providing political and economic background, this section informs the subsequent quantitative analysis and, in particular, the identification strategies that we employ.

3.A: Political Background

 

Responding to the decline of Africa’s economies in the post-independence period, many policy makers and political activists attributed it to politics.  Africa’s governments, they stressed, were authoritarian (Figure 1); not having to win the support of their people, they were free to impose policies that served their private interests[4] rather than the needs of the public.  The implication was clear: by changing Africa’s political institutions and rendering their governments accountable, reformers could alter their choice of policies, thereby promoting economic development.

 

The reformers consisted of political dissidents within Africa and persons abroad: political activists, of course, but, of more immediate relevance, technocrats in financial institutions.  The technocrats focused on the government’s management of the macro-economy.  Fiscal imbalances and lax monetary policies led, it was argued, to high rates of inflation and adverse trade balances, necessitating a search for loans from abroad.  They also criticized the manner in which governments intervened in markets, using trade policies to protect domestic industries from foreign competition and licensing requirements to protect them from domestic competition, while leaving farmers unsheltered.

 

As noted and stressed by Krueger, Schiff et al. (1992) and others (Ndulu and O'Connell 2007 and Anderson and Masters 2009), both the macro-economic policies and the interventionist measures adopted by Africa’s post-independence regimes resulted in urban bias.  When coupled with rigid exchange rates, the macroeconomic measures generated an appreciation of the domestic currency, which undermined the competitiveness of agricultural exports and lowered the price of foodstuffs imported from abroad.  While urban industries were sheltered from foreign competition, rural producers were not; and while firms could legally maintain monopolies in domestic markets, farms confronted legal monopsones, as governments restricted competitition for the crop as a way of reducing prices for consumers.  The mix of policies shifted relative prices against farmers. 

 

Given that agriculture production constituted a major portion of Africa’s exports, as they weakened incentives for farming, Africa’s governments also undermined incentives to generate earnings abroad.  To secure foreign exchange, they therefore increasingly turned to the World Bank and International Monetary Fund.  After repeated applications for funding from Africa’s debtor nations, these institutions demanded that they alter their policies and “governance structures.” 

 

As depicted in Table 1, institutional change began in French speaking West Africa:[5]  In February of 1990, in Benin, local reformers convened a national convention, which legalized opposition parties and called for open elections to fill public offices.  In response to events in Benin, the practice spread through neighboring states, then inland and southward, penetrating into Central and Southern Africa. In keeping with this account, in what follows, we employ indicators of macro-economic imbalances and relative prices as measures of public policy and the introduction of political competition as a measure of institutional change. 

3.B: Economic Background

 

In a recent paper, [] (2010) combined data from 44 countries over 46 years (1961-2007) to generate estimates of changes in total factor productivity in African agriculture.  In the initial years of independence, he found, total factor productivity dramatically declined.  In the early 1980s, however, it began to grow.  And by the early 2000s, its average annual rate of growth was over four times faster than it had been 25 years earlier.  His estimates suggest that the average rate of TFP growth in the baseline estimate is 0.97% per year, a figure that falls to 0.87% per year when we adjust for land quality and to 0.59% per year when we include adjustments for the quality of labor.  For our purposes, however, the key finding from [] (2010) from which we depart is the post-independence decline and subsequent rise of cross-country agricultural productivity growth in Africa. In late-century Africa, political institutions changed: governments that had once been authoritarian now had to compete for office.  The question therefore arises: Did this political change bear a systematic relationship to these changes in the performance of Africa’s rural economy?  We argue that it did, both directly and through its impact on government policies.

 

In the section that follows, we describe our data.  We then employ it to estimate the relationship between institutional reform, policy change, and the performance of Africa’s agrarian economies.

4. The Data

As our measure of economic performance, we employ []’s estimates of TFP growth in agriculture.  Using aggregate crop output figures for each country, and Africa-specific prices and PPP exchange rates,[6] [] derives his estimates from a semi-parametric specification of a  constant returns to scale Cobb-Douglass production function:

(1)   

where yi(t) is aggregate crop output for country i in year t, xij(t) is a vector of j conventional agricultural inputs (land, chemical fertilizer, tractors, and livestock), zij(t) are quality shifters associated with these inputs (average years of schooling to adjust labor quality, as well as rainfall and irrigated land share to adjust for the quality of land), pij(t) are other potential explanations for TFP growth (to include political competition), TD are annual time dummies, and CD are country dummies.  All variables are in logs, normalized by the size of the labor force in agriculture.

 

This semi-parametric specification effectively partials out the linear effects of the conventional inputs and country dummies, measuring TFP growth with a non-parametric kernel regression of output on the annual time dummies, .[7]  For arbitrarily small changes in time, the rate of TFP growth can be derived by differentiating :

 

(2)                

To derive the country-specific rates of agricultural TFP growth – the principal dependent variable in this paper -- we estimate equation (1) country-by-country.  The “baseline” estimates (shown in the cross-country aggregates in Figure 2) exclude the adjustments for input quality contained in the vector z.  We then re-estimate the function while adjusting for land quality (by controlling for the effect of annual rainfall and irrigated land share), and then re-estimate it once again while adjusting as well for labor quality (by controlling for average years of schooling).  While the estimate of TFP growth is reduced by the extent to which those additional variables “explain” the initial baseline estimate, the adjustments help to differentiate between productivity increases resulting from the use of improved inputs from those that result from increases in the efficiency with inputs are employed. 

 

 

As our measure of institutions, we make use of a scale that provides a measure of the degree of political competition that the incumbent chief executive faced when coming to office. [8]  For each country in each year, the scale assigns a number that indicates whether:

1 -- No executive exists.

2 – An executive exists but was not elected.

3 – The executive was elected, but was the sole candidate.

4 – The executive was elected, with multiple candidates competing for the office.

5 -- Multiple parties were also able to contest the executive elections.

6 -- Candidates from more than one party competed in executive elections, and the winner won more than 75% of the votes.

7 -- Candidates from more than one party competed in executive elections, and the winner won less than 75% of the vote.

 

We employ as our measure a dummy variable, named “electoral competition,” that takes the value 1 when the government is rated 6 or above and 0 otherwise.[9] 

 

When discussing public policy, we employ two measures.  One is the black market premium (BMP) for foreign exchange.  While the BMP is a direct measure of exchange rate misalignment, we follow Rodriguez and Rodrik (2001) who view BMP as a proxy for broader distortions in macroeconomic policy.  Researchers point to the misalignment of the exchange rate as a major disincentive to agricultural producers (e.g. Krueger, Schiff et al. 1992): overvaluation reduces the local currency earnings of those who produce cash crops for export and, in the absence of offsetting tariffs or quantitative restrictions, confers a price advantage to imported agricultural products. 

 

Our second policy indicator is the nominal rate of assistance to agricultural importables, a sectoral indicator of trade policy intervention, which we take from the World Bank’s database on Distortions to Agricultural Incentives (Anderson 2009).  When an ad valorem tariff is the sole policy intervention for good (x), the nominal rate of assistance for commodity x is:

(3)      

where tm  is tariff rate, E is the nominal exchange rate, and P is the dollar-denominated world price of the commodity.  This basic formula can be modified to incorporate additional distortions, such taxes and subsidies on domestic production of the relevant commodities.  The nominal rates of assistance for individual crops may be aggregated to form the nominal rate of assistance for agricultural importables (NRA_totm), which are typically foodgrains.  Food imports compete with the product of Africa’s famers for the domestic market.  NRA_totm thus determines the prices at which local producers can sell what they grow.  It also influences the distribution of income, as the lower prices that lower the fortunes of rural producers enhance the purchasing power of urban consumers.  It follows from our central line of reasoning that the enfranchisement of the rural majority should be associated with increased rates of nominal assistance for agricultural importables. 

 

Table 2 provides descriptive statistics for these data.

 

5. Analysis

5.A. Bivariate Relations

To motivate the analysis that follows, we introduce Figure 3, which distinguishes the time path of TFP growth rates in observations with and without electoral competition[10].  The TFP growth rate in settings characterized by electoral competition progressively diverges from the TFP growth rate in settings that lacked it.  On average, countries with electoral competition experienced agricultural TFP growth of 1.04% per year, while the average rate was only 0.48% per year in countries without.[11] 

 

We have suggested that changes in political incentives led to changes in the policy preferences of Africa’s governments.  Figure 4 therefore compares government policies in country-years with and without electoral competition.  Each panel in Figure 4 contains a box that depicts the portion of the observations of a variable that fall within the interquartile range, i.e. those whose values place them between the lower 25% and the upper 25% of the range of the values of the variable.  The horizontal lines within the boxes mark the variable’s median value.  The upper and lower horizontal lines laying outside the boxes mark the upper and lower values of the data.

 

The data suggest that governments headed by an executive chosen in a competitive election not only exercise greater fiscal and monetary restraint than do their authoritarian counterparts (as indicated by the virtual absence of black markets for their currencies) and intervene in markets in ways less likely to shift relative prices against farmers (as indicated by their relative rates of assistance); but also, that they spend more on agricultural research, secure higher levels of educational attainment, and pave a larger percentage of their roads.  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 act in ways that lower the costs, increase the earnings, and strengthen the incentives for farmers. 

 

In the estimates we develop below, we focus in particular on two of these policy variables: the black market premium on foreign exchange and the nominal rate of assistance to agricultural importables.  Figure 5 demonstrates the effect of controlling for the black market premium when estimating the time path of TFP growth.  The difference in the two time profiles can be viewed as the portion of TFP growth “explained” by the black market premium and suggests that macroeconomic policy distortions have “accounted for” a difference of nearly 30% in the quality-adjusted average rate of TFP growth.  Figure 6 applies this approach to the NRA_totm, which accounts for 16% of the estimated average growth rate of TFP.  In both cases, differences in the mean growth rates is significant at greater than the .10-level.

 

5.B. A Deeper Look: Difference-in-Differences

Probing deeper, we attempt to identify the causal effect of electoral competition on agricultural productivity growth.  In pursuit of that objective, we construct a difference-in-difference model in which electoral competition constitutes the treatment.  Given that the treatment occurs at different times in different countries, our model takes the form of a fixed effects regression with individual year dummies:

where  is the growth rate of agricultural productivity in country i in year t,  are time-invariant unobservable country effects,  are year dummies, X is a vector of observed covariates,  is a dummy equal to one for each country-year observation in which there is electoral competition, and δ is the causal effect of electoral competition on agricultural TFP growth (which we assume to be a constant).  We adjust all standard errors for clustering at the country level, in keeping with the cautions advocated by Bertrand, Duflo, and Mullainathan (2004) regarding serial correlation in difference-in-difference models.

 

Table 3 presents estimates of equation (4).  Column (1) presents a baseline specification that includes only country and year effects.  This specification indicates a statistically significant increase of 0.74 percentage points to agricultural TFP in settings with electoral competition.  The remaining columns of Table 3 test the robustness of this baseline finding.  Columns 2 through 5 introduce a set of covariates that include, respectively, the rural population share, a dummy for country-year observations with civil wars, an indicator of the extent of electoral competition in each countries geographic neighbors (averaged across neighbors and lagged by one year), and the combination of these covariates.  Rural population share can be interpreted as an indicator of development.  As expected, it enters negatively in column 2.  Its inclusion increases the point estimate for electoral competition, suggesting a relationship between the two that we explore below.  Civil wars were endemic in late century Africa, with 40% of countries experiencing at least one year of war between 1960 and 2000.  A dummy for each country-year of warfare enters negatively in column 3, but the estimated effect is statistically insignificant.  If electoral competition does indeed lead to improved policies and higher rates of agricultural productivity growth, its impact could conceivably spill across national boundaries; it would then influence the relationship between political reform and economic performance in neighboring states, introducing endogeneity bias.  Column 4 therefore controls for the lagged average of the degree of electoral competition in each country’s neighbors; and indeed the coefficient is positive.[12] 

 

A major threat to these estimates is that the effect may precede the treatment.[13]  As a robustness test against this threat, we therefore add country-specific time trends to our set of control variables, modifying equation (4) to read:

where  remains a country-specific intercept and  is a country-specific trend coefficient multiplying the time trend t.  Table 3 implements this robustness test in column 6, which includes all covariates.  This result is a small reduction in the point estimate for the effect of electoral competition on agricultural TFP growth, which remains statistically significant at the .05-level.

 

As an additional robustness test we seek to ensure that past treatment causes the current effects while future treatment does not, .In addressing this possibility, we follow Angrist and Pischke (2009) who invoke a form of Granger causality.  The relevant specification then becomes:

which allows for m lags (post-treatment effects) and q leads (anticipatory effect).  Figure 7 graphs the coefficient estimates of these post- and pre-treatment effects for m = q = 4 leads and lags surrounding the year in which each country transitioned into a system of competitive elections.  The results indicate no significant anticipatory effect on changes in agricultural productivity.  The difference between the mean coefficients before and after political transition is 0.56 percentage points, or approximately the magnitude reported in Table 3, column 6.  Note that Figure 7 also suggests the possibility that the impact of electoral competitiveness decays over time.

5.C. Mechanisms

Institutional change thus appears to bear a systematic relationship to economic performance in rural Africa.  The question remains: Through what mechanisms does this effect operate?  In keeping with our previous argument, we focus on government policies, starting with the nominal rate of assistance to agricultural importables (NRA_totm).

 

Table 4 presents our key results.  Column 1 reports the estimates derived from a random effects regression of NRA_totm on our dummy for electoral competition, controlling for country and year effects and using the largest sample available.[14]  Electoral competition has a substantial and statistically significant positive effect, supporting our hypothesis that electoral competition leads to policies favorable to Africa's food producing majority.  Allowing for country-specific trends in NRA_totm fails to alter this result.  The magnitude of these point estimates is approximately one-half of the sample standard deviation for NRA_totm, indicating a 17 percentage point increase in the rate of protection accorded import-competing food producers (roughly the difference between the 25th and 75th percentiles of the distribution).  In anticipation of the smaller sample available with added control variables (see below), column 3 repeats the estimation of column 2 but with the smaller sample.  The resulting loss of degrees of freedom slightly reduces the point estimate and increases its standard error such as to push the resulting t-statistic just beyond the range of statistical significance (P = 0.11).

 

While we believe the risk of reverse causality (in the sense that NRA_totm would cause electoral competitiveness)is minimal, we freely acknowledge the possible impact of excluded variables.  As suggested above, pressure from the donor community is a prime candidate for such a variable: it credibly could account for the co-variation of electoral competitiveness and policy support for domestic food producers.  We therefore introduce a dummy variable indicating whether a country in a given year was under any form of agreement with the IMF.  Adding this variable (column 4) while continuing to control for country-specific trends leaves the point estimate for the coefficient on electoral competition unchanged while slightly increasing its precision.  The indicator for IMF support itself is statistically insignificant in this specification. 

 

There is an additional concern, however: that IMF agreements are not randomly distributed across countries.  In column 5 we therefore estimate a two-stage model in which we, as do others (e.g. Easterly 2005), instrument for IMF agreements using each country's level of US military assistance and previous colonial status.[15]  The resulting coefficients (column 5) confirm the statistically significant relationship between electoral competition and NRA_totm, albeit with a somewhat smaller point estimate. 

 

If electoral competition in Africa increased the political influence of the rural majority, then the magnitude of this effect could reasonably be a function of the size of that population.  Investigating this possibility, we expand our specification to include an interaction term between electoral competition and rural population share.  We present these results in column 6 of Table 4, which evaluates the total effect of electoral competition on NRA_totm at the 25th, 50th, and 75th percentiles of the sample distribution.  At the lower level of rural population share, the effect of electoral competition is positive but not statistically different from zero.  The effect more than doubles (and is statistically different from zero) at the median level and increases by an additional 70% (to 0.13) at the 75th percentile.  That the point estimates nearly quadruple going from the 25th to the 75th percentiles of rural population share provides strongly suggestive evidence of the influence of rural enfranchisement on the policy preferences of governments.

 

As argued by Krueger, Schiff, and Valdes (1992), macroeconomic policies strongly impact upon agricultural incentives.  We have demonstrated above the bivariate relationship between the black market premia (BMP) for foreign exchange and the growth of agricultural TFP.  Here we examine the extent to which the introduction of electoral competition led to changes in the black market premium (Table 5).

 

Our initial specification includes only country and year effects along with the dummy variable for electoral competition.[16]  The resulting estimate suggests a significant and negative effect, on the order of one-fourth of the sample standard deviation, suggesting that electoral competition causes improved macroeconomic policies.  The inclusion of country-specific time trends in black market premia (column 2) reduces the point estimate, which nonetheless remains both economically and statistically significant.

 

As with our analysis of NRA_totm, we are concerned with potential of endogeneity arising from the exclusion of variables that may influence both political and macroeconomic policy reform.  And, once again, we are drawn to the potential impact of donor pressure.  In anticipation of the loss of degrees of freedom when we control for IMF agreements, we first repeat the estimation of column 2 in column 3, using the smaller sample.  The substantial reduction in available degrees of freedom results in a statistically insignificant coefficient for “electoral competition” and one that is of the “wrong” sign.  Instrumenting as before for the IMF dummy causes electoral competition to become statistically insignificant; and the re-introduction of country-specific trends renders the electoral competition dummy marginally significant effect and of the "wrong" sign.  While the impact of institutional reform on macro-economic policy is robust to the inclusion of country-specific time trends, it thus disappears when we attempt to purge the relationship of endogeneity bias.

 

Consistent with our central hypothesis, our difference-in-differences analysis thus demonstrates that agricultural TFP growth is significantly greater in settings with electoral competition.  In investigating the extent to which government policies may act as mechanisms through which electoral competition affects agricultural TFP growth, we found strong evidence that countries with electoral competition adopt trade policies that favor farming, but only suggestive evidence of an association between macroeconomic policy and electoral competition.  Table 6 summarizes these findings by regressing agricultural TFP growth rates against these mechanisms, with and without including electoral competition.  Should the inclusion of electoral competition eliminate what had previously been a significant effect of policy on TFP growth, this would indicate that the policy variable provides a mechanism linking electoral competition to economic perfromance.  Consistent with our difference-in-differences analysis, we find in the first two columns of Table 6 that NRA_totm is significant only in the absence of electoral competition, but that the point estimates for black market premium are not affected by the inclusion of electoral competition.  In both cases, electoral competition itself remains a significant determinant of agricultural TFP growth.

 

  1. 6.     Conclusions

As stressed by Persson and Tabellini (2008), efforts to estimate the relationship between political institutions and economic performance run afoul of two major problems: heterogeneity and invariance.  By drawing our cases from a single region – Africa – we address the first difficulty; Africa’s economies lodge among the ranks of the poor and agrarian.  Politically, they are notably unstable.  It is precisely because of the frequency with which institutions change in Africa, however, that the continent affords us variation in political institutions that elude those wedded to the use of global samples.  Focusing on late century Africa, we take advantage of the data generated by efforts at political reform to study the relationship between democracy and development.

 

In doing so, we focus on what many regard as the core challenge to Africa’s economic development: the performance of its rural economy. Building on a recent analysis of agricultural productivity growth in Africa, we employ a difference-in difference approach and conclude that the introduction of electoral competition was systematically related to an increase of between 0.5 and 1.0 percentage points in the growth rate of total factor productivity in African agriculture.  We find that the transition to electoral competition led to significant increases in the rate of protection offered Africa’s food-producers.  The magnitude of this effect appears to have been greater in settings with larger rural majorities.  Less persuasively, we also found evidence that electoral competition led to improved macroeconomic policy, thus leading to higher domestic prices for food producers.  Taken together, the evidence suggests that the search for rural majorities led to changes in government policies, which in turn strengthened the incentives for farming in Africa.


Table 1. The Diffusion of Political Reform

Country

Date

Duration

Election

Outcome: Incumbent

Month

F&F?

Ousted

Retained

Benin

Feb-90

1 week

Feb-91

yes

 

 

 

 

Mar-96

yes

 

Congo

Feb-91

3 months

Aug-92

yes

 

Gabon

Mar-90

3 weeks

Dec-93

no

 

Mali

Jul-91

2 weeks

Apr-92

yes

 

Niger

Jul-91

6 weeks

Feb-93

yes

 

Burkina Faso

Aug-91

2 months

Dec-91

no

 

Ghana

Aug-91

7 months

Dec-92

yes

 

Togo

Aug-91

1 month

Aug-93

no

 

Zaire

Aug-91

1 year

--

--

 

 

CAR

Oct-91

2 months

Aug-92

yes

 

Chad

Jan-93

3 months

Jun-96

no

 

 


Table 2.  Variables and Descriptive Statistics

Variable

Obs

Mean

Std. Dev.

Min

Max

Source:

 

 

 

 

 

 

 

Agricultural TFP Growth

1494

0.614

2.117

-7.694

8.247

[] (2010)

Dummy=1 if Exec. Index of Electoral Competition >6

1460

0.427

0.495

0.000

1.000

Beck and Clarke (2009)

Neighbors' Executive Index of Electoral Competition

1230

4.289

1.586

1.500

7.000

Based on Beck & Clarke (2009)

Relative Rate of Assistance (RRA)

642

-0.279

0.299

-0.946

1.295

Anderson and Valenzuela (2008)

Black Market Premium on Foreign Exchange

1321

1.361

3.436

-6.908

6.122

World Devt Indicators (2009)

Civil War dummy

2162

0.166

0.372

0.000

1.000

Sambanis and Doyle (2006)

Rural Population Share

2064

71.713

16.410

12.700

97.960

World Devt Indicators (2009)

 

Countries for which we have estimates of agricultural TFP growth (boldface indicates the existence of data for RRA for that country):  Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Democratic Republic of Congo, Côte d'Ivoire, Equatorial Guinea, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Malawi, Mali, Mauritania, Mauritius, Mozambique, Niger, Nigeria, Rwanda, Senegal, Seychelles, Sierra Leone, Somalia, South Africa, Swaziland, Tanzania, Togo, Uganda, Zimbabwe.

Table 3.  The Effect of Electoral Competition on Agricultural Productivity Growth

 

(1)

(2)

(3)

(4)

(5)

(6)

VARIABLES

Dependent Variable:  Growth Rate of Agricultural TFP

Electoral Comp dummy

0.744*

0.911**

0.780*

0.655

0.828**

0.548**

 

(0.432)

(0.362)

(0.433)

(0.457)

(0.362)

(0.218)

Rural Pop. Shr.

 

-0.159***

 

 

-0.158***

-0.0552

 

 

(0.0373)

 

 

(0.0354)

(0.198)

Civil War dummy

 

 

-0.452

 

-0.163

-0.226

 

 

 

(0.493)

 

(0.394)

(0.150)

Neighbors’ electoral comp (t-1)

 

 

 

0.294*

0.313*

0.193

 

 

 

 

(0.155)

(0.155)

(0.125)

Constant

0.0377

10.48***

0.0800

-1.598*

8.567***

53.26

 

(0.552)

(2.449)

(0.593)

(0.881)

(2.402)

(260.3)

 

 

 

 

 

 

 

Observations

635

635

635

635

635

635

R-squared

0.126

0.308

0.139

0.150

0.336

0.677

Number of countries

27

27

27

27

27

27

Std Errors Clustered at Country-level

YES

YES

YES

YES

YES

YES

Country FE

YES

YES

YES

YES

YES

YES

Year FE

YES

YES

YES

YES

YES

YES

Country-Specific Trends

NO

NO

NO

NO

NO

YES

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

 

Table 4.  Effect of Electoral Competition on Nominal Rate of Assistance to Agricultural Importables

Dependent Variable: Nominal Rate of Assistance to Agricultural Importables

 

(1)

(2)

(3)

(4)

(5)

(6)

VARIABLES

RE

RE

RE

RE

RE-2SLS

RE-2SLS

Electoral Comp. dummy

0.170***

0.178***

0.102

0.101*

0.0781*

-0.216

 

(0.0570)

(0.0606)

(0.0637)

(0.0615)

(0.0428)

(0.269)

Under IMF Agreement

 

 

 

-0.0033

-0.156*

-0.224**

 

 

 

 

(0.0500)

(0.0903)

(0.103)

Rural Pop. Share

 

 

 

 

 

0.0259**

 

 

 

 

 

 

(0.0125)

Electoral Comp x Rural Pop. Shr.

 

 

 

 

 

0.00408

 

 

 

 

 

 

(0.00353)

Constant

0.000218

1.822

-16.31**

-16.49**

 

 

 

(0.101)

(5.131)

(7.617)

(7.668)

 

 

Effect of Electoral Competitiveness evaluated with rural pop. share at:

 

 

 

 

 

25th pctl (61.5%)

 

 

 

0.035

(0.065)

50th pctl   (72%)

 

 

 

 

0.077*

(0.054)

75th pctl (84.5%)

 

 

 

0.129**

(0.054)

 

 

 

 

 

 

 

Observations

548

548

347

347

347

347

Number of ccode

18

18

18

18

18

18

Std Errors Clustered at Country-level

YES

YES

YES

YES

.

.

Country FE

YES

YES

YES

YES

YES

YES

Year FE

YES

YES

YES

YES

YES

YES

Country-Specific Trends

NO

YES

YES

YES

YES

YES

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

P = 0.11

 

Table 5.  Effect of Electoral Competition on Black Market Premium

Dependent Variable:  log Black Market Premium on Foreign Exchange

 

(1)

(2)

(3)

(4)

(5)

(6)

VARIABLES

RE

RE

RE

RE

RE-2SLS

RE-2SLS

 

 

 

 

 

 

 

Electoral Competition    dummy

-1.054***

-0.842*

0.513

0.395

-0.0797

0.601*

(0.353)

(0.443)

(0.434)

(0.422)

(0.317)

(0.347)

Under IMF Agreement

 

 

 

-0.721**

-0.377

0.544

 

 

 

 

(0.364)

(1.082)

(0.939)

Constant

3.532***

620.4***

696.4***

649.6***

3.837***

 

 

(0.435)

(50.83)

(87.12)

(85.15)

(0.860)

 

 

 

 

 

 

 

 

Observations

686

686

390

390

390

390

Number of Countries

33

33

31

31

31

31

Std Errors Clustered at Country-level

YES

YES

YES

YES

.

.

Country FE

YES

YES

YES

YES

YES

YES

Year FE

YES

YES

YES

YES

YES

YES

Country-Specific Trends

NO

YES

YES

YES

NO

YES

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

 

 

Table 6.  NRA and Black Market Premium as Mechanisms for Electoral Competition

Dependent Variable:  Agricultural TFP Growth

 

(1)

(2)

(3)

(4)

VARIABLES

RE

RE

RE

RE

 

 

 

 

 

NRA_totm

0.943*

0.707

 

 

 

(0.524)

(0.542)

 

 

Electoral Comp.

 

0.855***

 

0.918***

 

 

(0.324)

 

(0.325)

Log Black Mkt Premium

 

 

-0.273*

-0.275**

 

 

 

(0.141)

(0.122)

Constant

1.118

0.254

1.840**

1.174***

 

(0.816)

(1.011)

(0.929)

(0.414)

 

 

 

 

 

Observations

258

258

258

258

Number of countries

13

13

13

13

Std Errors Clustered at Country-level

YES

YES

YES

YES

Country effects

YES

YES

YES

YES

Year dummies

YES

YES

YES

YES

 

Robust standard errors (clustered at country level) in parentheses

*** p<0.01, ** p<0.05, * p<0.1

 

 

Figure 1: Index of Political Competition

Source:  Beck and Clarke (2009); Harvard University Africa Research Project (http://africa.gov.harvard.edu/).

 

 

Figure 2.  Agricultural TFP Growth Rates Adjusted for Input Quality

 

Source:  [] (2010)

 

 

 

 

 

Figure 3.  Agricultural TFP Growth Profile for Country-Years With and Without Electoral Competition

 

 

 

Figure 4

 

Figure 5.  The Effect of Black Market Premium on Agricultural TFP Growth Rates

 

 

Source: [] (2010)

 

 

Figure 6.  Effect of Nominal Rate of Assistance to Agricultural Importables on Agricultural TFP Growth

 

 

 

Figure 7.  Coefficients on Pre- and Post-Treatment Effects of Electoral Competition on Agricultural TFP Growth

 

 

 

 

 References

 

Acemoglu, D., S. Johnson, et al. (2008). "Income and Democracy." American Ecoonomic Review 98(3): 804-842.

 

Anderson, K., and E. Valenzuela (2008). Estimates of Distortions to Agricultural Incentives, 1955 to 2007, core database at http://www.worldbank.org/agdistortions

 

Angrist, J. D. and J.-S. Pischke (2008). Mostly Harmless Economics. Princeton, Princeton University Press.

 

Barro, R. (1996). "Democracy and Growth." Journal of Economic Growth 1: 1-27.

 

Beck, T., G. Clarke, et al. (2001). "New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions." World Bank Economic Review. (updated, 2009)

 

Bertrand, M., E. Duflo, et al. (2004). "How Much Should We Trust Differences-in-Differences Estimates?" Quarterly Journal of Economics 119(1): 241-275.

 

Blundell, R. and S. Bond (1998). "Initial conditions and moment restrictions in dynamic panel data models." Journal of Econometrics 87(1): 115-143.

 

Boix, C. and S. C. Stokes (2003). "Endogenous Democratization." World Politics 55(4): 517-549.

 

Chenery, H. B. and L. J. Taylor (1968). "Development Patterns: Among Countries and Over Time." Review of Economics and Statistics(November): 391-416.

 

Collier, P. and J. W. Gunning (1999). "Why Has Africa Grown Slowly?" Journal of Economic Perspectives 13(3): 3-22.

 

Dahl, R. (1971). Polyarchy. New Haven, Yale University Press.

 

Dunning, T. (2004). "Conditioning the Effects of Aid: Cold War Politics, Donor Credibility, and Democracy in Africa." International Organization 58(2): 409-423.

 

Easterly, W. (2005). "What did structural adjustment adjust?: The association of policies and growth with repeated IMF and World Bank adjustment loans,." Journal of Development Economics 76(1): 1-22.

 

Easterly, W. and R. Levine (1997). "Africa's Growth Tragedy: Policies and Ethnic Divisions." Quarterly Journal of Economics 112(4): 1203-1250.

 

Ferree, K., C. Gibson, et al. (2009). Explaining the African Vote. San Diego, Department of Political Science, University of California.

 

Giavazzi, F. and G. Tabelinni (2005). "Economic and Political Liberalizations." Journal of Monetary Economics 52: 1297-1330.

 

Glaeser, E., R. La Porta, et al. (2004). "Do Institutions Cause Growth?" Joural of Economic Growth 9: 271-304.

 

Keefer, P. (2010). The Ethnicity Distraction? Political Credibility and Partisan Performance in Africa. Policy Research Working Paper 5236. Washington DC, The World Bank.

 

Krueger, A. O., M. Schiff, et al., Eds. (1992). The Political Economy of Agricultural Pricing Policies, 5 vols. Baltimore, Published for the World Bank by Johns Hopkins University Press.

 

Kudamatsu, M. (2007). Has Democratization Reduced Infant Mortality in Sub-Saharan Africa? Evidence from Micro Data. SSRN Discussion Paper No. 685. Stockholm, Stockholm University - Institute for International Economic Studies.

 

Kuznets, S. (1966). Modern Economic Growth. New Haven and London, Yale University Press.

 

Lipset, S. M. (1959). "Some Social Requisits of Democrcy: Economic Development and Political Legitimacy." American Political Science Review 53: 69-105.

 

Lipton, M. (1977). Urban Bias. London, Temple Smith.

 

Little, I. M. D., T. Scitovsky, et al. (1970). Industry and Trade in Some Developing Countries. Oxford, Oxford University Press.

 

Mattes, R. and J. Piombo (1999). "Opposition Parties and  the Voters in South Africa's General Election of 1999." Democratization 8(3): 101-128.

 

Ndulu, B. (2007). Chapter 9: The Evolution of Global Development Paradigms and their Influence on African Economic  Development. The Political Economy of Economic Growth in Africa, 1960-2000. B. Ndulu, P. Collier,et al. Cambridge, University of Cambridge Press.

 

Ndulu, B. J. and S. A. O'Connell (2007). Chapter 1: Policy Plus: African Growth Performance 1960-2000. The Political Economy of Economic Growth in Africa, 1960-2000. B. J. Ndulu, P. Collier, et al. Cambridge, Cambridge University Press.

 

Olson, M. (1971). The Logic of Collective Action. New York, Schoken Books.

 

Olson, M. (1985). "Space, Agriculture, and Organization." American Journal of Agricultural Economics 67: 928-937.

 

Persson, T. and G. Tabellini (2008). The Growth Effect of Democarcy. Institutions and Economic Performance. E. Helpman. Princeton, Princeton Universty Press: 544-586.

 

Posner, D. N. and D. J. Simon (2002). "Economic Conditions and Incumbent Support in Africa's New Democracies." Comparative Political Studies 35(3): 313-336.

 

Przeworski, A., M. Alvarez, et al. (2000). Democracy and Development. New York, Cambridge University Press.

 

Rodriguez, F. and D. Rodrik, 2001. "Trade Policy and Economic Growth: A Skeptic's Guide to the Cross-National Evidence," NBER Chapters, in: NBER Macroeconomics Annual 2000, Volume 15, pages 261-338 National Bureau of Economic Research, Inc.

 

Sambanis, N., and M. Doyle. 2006.  Making War and Building Peace: United Nations Peace Operations. Princeton, NJ: Princeton University Press.

 

Schumpeter, J. A. (1950). Capitalism, Socialism and Democracy. New York, Harpers and Row.

 

Stasavage, D. (2005). "Democracy and Education Spending  in Africa." American Political Science Review 49(2): 343-358.

 

Stokes, S. C. (2005). "Perverse Accountability: A Formal Model of Machine Politics and Evidence from Argentina." American Political Science Review 99(3): 315-325.

 

Varshney, A. (1995). Democracy, Development and the Countryside. Cambridge, Cambridge University Press.

 

Varshney, A. (1995). Democracy, Development and the Countryside. Cambridge, Cambridge University Press.

 

Vicente, P. C. and L. Wantchekon (2009). "Clientelism and Vote Buying: Lessons from Field Experiments in Africa." Oxford Review of Economic Policy 25(2): 292-305.

 

World Bank. (1981). Accelerated Development in Sub-Saharan Africa: An Agenda for Action. Washington DC, The World Bank.

 

World Bank. (2008). World Development Indicators. Washington DC, The World Bank.

 

Yatchew, A. 2003. Semiparametric Regression for the Applied Econometrician. Cambridge: Cambridge University Press.

 

 


[1]  Reference masked.

[2] Economic conditions that may be beneficial to the bond holder may be regarded unfavorably by the marginally or un-employed.

[3] That is: being a Bemba-speaker (as was the incumbent) or not.

[4] Ideological as well as personal, as eloquently argued by Ndulu, B. (2007). Chapter 9: The Evolution of Global Development Paradigms and their Influence on African Economic  Development. The Political Economy of Economic Growth in Africa, 1960-2000. B. Ndulu, P. Collier, et al. Cambridge, University of Cambridge Press.

[5] In 1989 in French-speaking Africa, many drew inspiration from the 1989 bicentennial of the French Revolution.  They saw themselves as continuing the struggle for the rights of citizens, launched in Paris two hundred years before.

[6] [] (2010) constructs these aggregates from crop-specific output data published by the Food and Agricultural Organization of the UN.  Other studies simply employ the FAO’s pre-constructed output aggregates, which are based on global prices and exchange rates.  []’s estimates thus more closely reflect the circumstances actually faced by Africa’s farmers.

[7] Yatchew (2003) provides comprehensive detail on semi-parametric regression.

[8] The measure is taken from the World Bank’s Data Base of Political Institutions: Beck, T., G. Clarke, et al. (2001). "New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions." World Bank Economic Review.

[9] Others we term “non-competitive” or, more loosely, “authoritarian”. 

[10] Net of adjustments for input quality.

[11] These averages are statistically different in a two-sided t-test (P = 0.0014).

[12] Reference masked.

[13] Note that one cannot judge this sequence from Figure 4, which is aggregated across all countries in the sample.  Analyzing the sequence of treatment and effect requires country-level disaggregation.

[14] Our use of random rather than fixed effects models in Table 4 is largely pragmatic.  Random effects models are more efficient than fixed effects models, and exploit both "within" and "between" variation in panel data.  Yet, random effects specifications also require the assumption that the time-invariant country effect is uncorrelated with the other regressors.  This latter assumption is not justified in our sample, introducing the potential for bias.  However, most of the identifying variation in NRA_totm is cross sectional, and fixed effects models performed poorly in this application.  An additional pragmatic consideration favoring random effects is that it allows us to retain time-invariant regressors, the need for which is described below.  Nonetheless, the poor performance of fixed effects models must count against our argument.

[15] That the latter of these two instrumental variables is time invariant reinforces our use of random effects estimators, as such instruments would drop out of fixed effects models.

[16] We restrict this sample to exclude countries of the CFA zone, as they do not control their own foreign-exchange regimes and electoral competition should have no effect.