Journal of Economic Growth, 4: 81–111 (March 1999) c 1999 Kluwer Academic Publishers, Boston. ° Economic Growth and Political Regimes J. BENSON DURHAM Columbia University, New York, NY 10027 The absence of continuous regime type measures that focus on institutions rather than outcomes besets studies on whether democratic or authoritarian regimes grow faster. Additional shortcomings include the failure to consider development stages and the erroneous endogenous specification of regimes. Given panel data on 105 countries from 1960 to 1989, the effective party/constitutional framework measure does not correlate with growth or investment in the total sample. But considering development levels, some evidence indicates that discretion decreases growth in advanced areas, and, contrary to theory, inhibits investment in poorer countries. Also, single-party dictatorships have higher investment ratios but do not grow faster than party-less regimes. Keywords: growth, investment, regimes JEL classification: 040, 057 1. Introduction Do dictatorships or democracies better promote economic growth? Given strong normative preferences for both democracy and development, this political economy question receives much deserved attention among scholars as well as policy entrepreneurs. After considerable debate and econometric evidence, however dubious, an influential literature review concludes that the inquiry “is wide open for reflection and research” (Przeworski and Limongi, 1993, p. 66). To begin, one must first inquire how regimes, which refer to the methods politicians must use to gain and maintain control of the state, affect growth. The key causal institutional characteristic of regimes is the degree of policymaker discretion or freedom of action that such methods give sovereigns. Some social scientists recommend extensive sovereign discretion, while others endorse limited government. Discretion is the crux of both arguments. Persistent problems in econometric studies warrant additional evaluation. First, and perhaps most important, the literature does not consider continuous proxies for regime type that focus more acutely on institutions as opposed to outcomes. Given that arguments rest on the continuous underlying concept of policymaker discretion, proxies for political regime type should also be continuous and thereby distinguish among democracies and dictatorships, hitherto a residual econometric category. Furthermore, considering the practical policy implications of the debate, measures should capture a continuum of discretion with respect to meaningfully serviceable institutional concepts, such as the number of effective political parties and constitutional frameworks, rather than outcome proxies that economists routinely employ. Therefore, this article presents an alternative regime type proxy to com- 82 DURHAM plement existing measures that is continuous rather than discrete and objective rather than subjective. But before applying a measure that focuses more acutely on institutions, two additional issues germane to this econometric question should be thoroughly outlined. First, existing tests do not explicitly evaluate the tacit consensus that dictatorship is more useful in less developed nations, while democracy is more suitable in advanced countries—a critical addendum that posits ambiguous results for universal samples. Therefore, unlike previous studies, the sample is divided according to levels of GDP per capita to approximate the developmental circumstances under which discretion is either beneficial or debilitating. Second, some social scientists erroneously suggest that the inquiry is intractably fraught with simultaneity and/or survival bias. Rather, given the routine distinction between the level and rate of economic development, the estimation problem is collinearity between initial GDP per capita and regime type. The remainder of Section 1 reviews the two competing perspectives. Section 2 outlines the effective party/constitutional framework measure. Section 3 discusses additional econometric issues—namely, development stages and simultaneity and/or survivor bias, which must be addressed before econometric estimation. Section 4 presents the econometric results for growth and investment in the total sample, across eight per capita GDP breaks, and exclusively in authoritarian regimes. Section 5 concludes. 1.1. The Authoritarian View Authoritarian regimes bestow considerable discretion on sovereigns and are by definition not susceptible to the “political business cycle.” In Kurth’s estimation, dictatorships more effectively stimulate growth and investment by suppressing labor unions, wages, and consumer demand—very unpopular measures (1979, p. 334; see also March, 1988; Sirowy and Inkeles, 1990). Put simply, democratic governments that wish to maximize tenure must respond to popular demands for spending and consumption. Assuming considerable sacrifices associated with capital accumulation and inherently myopic voters, political parties simply cannot simultaneously advocate sacrifices and win elections (Rao, 1984–85, p. 75; see also Weede, 1983, p. 36). Gershenkron’s early contributions are also illustrative and particularly relevant to historical development stages. Writing in contrast to the Marxist view that established industrial countries foreshadow development for less developed areas, he argues that “the advantages of backwardness,” the combination of cheap labor and technological backlogs, can accelerate development in poorer countries (1962, p. 8). Therefore, the rate of development differs across cases, and the political consequences are clear—namely, the greater the initial backwardness, the greater the required state intervention necessary to affect industrialization. Extensive state intervention, in turn, instinctively relates to authoritarian government. Effective policies require great sacrifice, and policymakers must have sufficient discretion to command such (myopic) hardship from the subject polity. An important subtlety to this view is that authoritarian rule is perhaps a necessary but certainly not a sufficient condition for industrialization. That dictators do not always extol discipline is not lost on Gerschenkron (1962, p. 20). Seemingly, regimes must be strong ECONOMIC GROWTH AND POLITICAL REGIMES 83 enough to affect investment, yet that very strength can potentially inhibit resource allocation and accumulation. This foreshadows the alternative view that democracy better fosters development. 1.2. The Democratic View Unlike dictatorships, democracies limit sovereign discretion and thereby more effectively promote economic growth because they solve the “credible commitment problem.” According to North and Weingast, if subjects suspect that sovereigns might alter property rights to extract rents, they will have less incentive to invest. Therefore, sovereigns cannot merely establish rights: they must also make subjects believe ex ante that they will not renege on their commitments (1989, p. 803). In contrast to authoritarian regimes, democracies check arbitrary rule, perhaps more effectively limit corruption, and thereby preserve established property rights. As Scully suggests, legal uncertainties increase transaction costs, and capital accumulation increases with more secure legal claims on property and income (1988, p. 653; see also Pourgerami, 1991, pp. 189, 194; Kohli, 1986, p. 165; Wittman, 1989, pp. 1397, 1411). Przeworski and Limongi reject this logic and consider the argument that democracy protects property rights a “recent” and “far-fetched invention” (1993, p. 52). They list Mackintosh, Ricardo, Macaulay, and Marx, who all suggest that extending suffrage to lower classes threatens private property. One might add Laske to this list, but the recent nature of the assertion that democracy ameliorates the credible commitment problem does not make it wrong. While Przeworski and Limongi argue that North does not directly relate property rights to democracy, Olson (1993), whom the authors neglect in this context, persuasively links the “state of nature,” sovereigns, property rights, and democracy. Very briefly, Olson argues that distortionary taxes are lower in democracies compared to dictatorships. In democracies, both the incumbent and the majority receive wealth transfers, as such governments redistribute some receipts to the constituents, while autocrats simply indulge themselves.1 Coupled with the succession problem in dictatorships that supposedly debilitates long-run commitments, democracies are therefore more efficient because at least some portion of the polity, the majority, also receives rents. Democratic consent limits sovereign discretion, and therefore policymakers spread wealth beyond (presumably kleptocratic) sovereigns to subjects.2 2. Toward a Continuous Regime Type Measure with Institutional Referents However efficient alternative regimes, arguments for either democracy or dictatorship ultimately rest on notions about sovereign constraint. The substance of policy notwithstanding, regimes should either give sovereigns the freedom of action to take benevolent measures or restrain them from malevolent actions. This central concept of discretion is a continuous and not a dichotomous variable, as many studies dubiously conceive. Also, the question has considerable ex ante proscriptive relevance: considering the ubiquitous preference for development, should new regimes in Congo, Indonesia, and Nigeria, 84 DURHAM for example, facilitate or inhibit executive action? Unfortunately, existing measures used in econometric studies do not adequately capture the continuous nature of executive discretion and objectively measurable institutions that distinguish regimes. Of course, existing scales, which influential studies such as Kormendi and Meguire (1985), Grier and Tullock (1989), and Barro (1996) employ, are ordinal values in raw form, but these ex post measures do not isolate specific institutions or address key constitutional choices relevant to policy entrepreneurs. 2.1. Common Econometric Measures To begin a brief review of common measures, Jaggers and Gurr (1995) provide a useful discussion of the plethora of quantitative expressions of regime type. These include data sets such as their own that are based on subjective criteria, including historical monographs, and more objective data sets such as Vanhanen’s, which relies exclusively on electoral data.3 Briefly, Jaggers and Gurr’s Polity III measures employ nine indicators of “institutionalized authority characteristics.” These indicators, in turn, endeavor to capture three concepts—“the influence dimensions of authority,” “the recruitment of chief executives,” and “government structure” (Jaggers and Gurr, p. 470). Another measure of regime type, perhaps the most popular, is Gastil’s Freedom House dual indices of political and civil rights. Political rights refer to the “capacity of individuals to participate freely and effectively in the selection of policymakers and binding decision rules affecting the national, regional, and local community.” Civil rights “are defined by the freedom of individuals to develop views, institutions and personal autonomy apart from the state” (Jaggers and Gurr, p. 474). Turning to validity and estimation, like Jaggers and Gurr’s, the Gastil indicators are inherently subjective, and time-series validity of these indices is highly questionable. Even the Freedom House judges readily admit that “changes in information and judgment since 1973 make many ratings not strictly comparable from year to year” (Gastil, 1987, p. 53).4 Validity aside, one might employ the ordinal forms to capture the degrees of executive discretion, but the unit changes in both variables have little substantive meaning. As Clague et al. note, the indices capture outcomes of processes rather than define specific features of a democratic political system (1996, p. 251). They do not refer to the institutions involved in, to borrow DiPalma’s (1990) phrase, “crafting democracies” ex ante. Few economists or political scientists have an intuitive understanding of how, say, a country with a measure of “1” differs from another with a measure of “5” on the twenty-one-point Polity III scale. More important, policy entrepreneurs cannot readily replicate such outcomes in, say, Congo, Indonesia, or Nigeria. 2.2. An Alternative Proxy: The Effective Party/Constitutional Framework Measure Given these practical shortcomings of common outcome measures, a useful alternative measure of regime type should capture more servicable concepts that relate to the critical notion of discretion. Furthermore, such a proxy should be continuous rather than discrete without losing its intuitive appeal and empirical relevance. This section outlines a measure ECONOMIC GROWTH AND POLITICAL REGIMES 85 that endeavors toward these ends. The alternative incorporates the number of effective political parties in government and the position of the executive vis-à-vis legislative offices. 2.2.1. Determination of the Existence and Number of Effective Parties The number of “effective” political parties in government constrains policymakers in two respects, within and without the governing party. Both internal policy discipline and external party competition constrain policymakers. The first obtains if a party exists in government. The second obtains if competitive elections allow additional parties in government. The fewer parties in government, the more discretion the regime has in policymaking. Before one can determine the “effective” number, one must define “political party.”5 This stipulative definition draws from Ranney (1968), Duverger (1954), LaPalombara and Weiner (1966), Von Der Mehden (1964), and Huntington (1968) and includes three essential elements. First, a party refers to an organized group that describes itself as such and seeks to control the state. Whoever occupies the executive office in government must be a member of such a party or the regime is party-less.6 Second, a party must compete in elections, competitive or otherwise. Elections, even noncompetitive contests, are critical because such practice indicates that the party attempts to channel or mobilize the polity, which Huntington (1968) emphasizes in the context of political development. Third, a party must exhibit organizational continuity. That is, the life span of the party does not depend on its founders. The party must either continue after the demise of its founders or an effective change in leadership must occur.7 This consideration measures the relative institutionalization of parties and, especially among authoritarian regimes, distinguishes them from more traditional political organizations. The stipulative definition applies irrespective of the number of effective parties in the regime. Also, given the emphasis on economic policymaking, the definition refers exclusively to the “ ‘party-in-the-government’—persons holding public legislative and executive offices who sometimes act with solidarity on certain matters” (Ranney, 1968, p. 154). Recall that proponents of the authoritarian model suggest that discretion is beneficial. If ideal parties mobilize or channel political sentiment from the entire polity, then the more embedded party would similarly resist unpopular measures for growth and investment. Thus, policymakers in one-party dictatorships should enjoy less discretion than their counterparts in party-less regimes. As Huntington suggests, “Patriot Kings” and “conservative nonroyal leaders” like Sarit, Ayub Khan, Franco, or Rhee view political parties as divisive because they challenge authority and complicate unification and modernization. More important, the bureaucracy that executes economic policy views political parties suspiciously because they introduce irrational corruption into the pursuit of unproblematic goals, such as growth (Huntington, 1968, pp. 403–404). Thus, even in an autocratic context, party discipline constrains dictators and limits their ability to pursue unpopular measures.8 Considering the general authoritarian view that discretion is beneficial, single-party dictatorships should grow more slowly than their party-less counterparts. Conversely, assuming the general democratic view that discretion is malevolent, single-party regimes should grow faster.9 Given the existence of a party, the second step is to determine the number of “effective” parties in government, which produces subtypes of democracies. Following Laakso and Taagepera, who do not apply their measure to economic outcomes, the “effective number 86 DURHAM of parties is the number of hypothetical equal-size parties that would have the same total effect on fractionalization of the system as have the actual parties of unequal size” (Laakso and Taageper, 1979, p. 4). The effective number N is the inverse of the sum of the squared fractured seat shares for each party in government.10 That is, N = Pn 1 l=1 Pl2 , (1) where N is the number of effective parties, and Pl represents the share of the total number of parliamentary seats that the lth party holds. Therefore, if the share of seats is equal among all parties, then N equals the actual number of parties n. If, in the authoritarian case, one party competes in a noncompetitive election and seats are distributed accordingly, then N equals one, the actual number of parties. 2.2.2. Incorporation of the Constitutional Framework With respect to policymaking discretion, in addition to the raw number of effective legislative parties, the constitutional framework in which those parties exist is very important. Accordingly, the effective party/constitutional framework proxy modifies Laakso and Taageper’s original formulation (1) to incorporate the relative position of the legislative vis-à-vis the executive branch, which is a critical locus of economic policy implementation if not formulation. This modification endeavors to more comprehensively capture governmental discretion. The distinction isolates “parliamentary” from other categories such as “presidential,” “semipresidential,” “mixed,” and “authoritarian” regimes. In Stepan and Skach’s terms, a “pure parliamentary regime” is a system of mutual dependence, where parliamentary majorities must support executive power, and the executive office can dissolve the legislature and call elections (such as in Italy, Japan, Jamaica, or the United Kingdom). On the other hand, a “pure presidential regime” is a system of mutual independence, where both the executive and legislative powers have fixed electoral mandates that are the sole sources of legitimacy (such as in Columbia, Costa Rica, the United States, or Venezuela) (Stepan and Skach, 1993, pp. 3–4, emphasis added). Similarly, unelected dictators, who may or may not permit the existence of a multiparty parliament, occupy the executive branch, and the system exhibits exclusive independence. In short, the raw number of effective parties captures discretion for parliamentary regimes, but the figure adjusts for nonparliamentary categories because the executive branch in such cases typically accommodates only one party. That is, given that “legitimacy” exclusively derives from the legislature in a parliamentary regime, the raw effective parties measure (1) is the proxy. But for systems of mutual dependence or exclusive independence the measure is the mean of the number of parties across the executive and legislative branches, as in Pn1 N= l=1 Pl2 + Pn1 e=1 2 Pe2 , (2) where Pe , the number of effective political parties in the executive branch, is equal to one.11 ECONOMIC GROWTH AND POLITICAL REGIMES 2.3. 87 Some Advantages and Disadvantages of the Effective Party/Constitutional Framework Measure This section briefly outlines the correlation matrix of alternative measures, notes more specific examples with respect to Table 1, and discusses the relative appropriateness of outcome versus institutional measures with respect to this econometric issue. In addition to the useful ordinal distinctions among democracies and dictatorships, which common outcome alternatives do not produce, the effective party/constitutional framework measure complements common proxies because it more directly relates to the institutional characteristics of regimes. Such is not to argue that outcome based proxies have no purpose. Rather, an alternative that focuses more acutely on institutions should more directly address the debate’s applicable implications. With respect to the reliability of the measure, the alternative proxy correlates comparatively highly with more common measures. The effective party/constitutional framework measure correlates closely with both the Polity III (.727) and Gastil (.720) measures, but not as closely as the Polity III and Gastil indices (.918).12 Turning to examples listed in Table 1, very few cases diverge considerably in rank. For example, the fifteen countries with the greatest number of effective parties are all clearly classified as democratic according to both the Gastil and Polity III indices.13 Conversely, the thirty-nine single-party and party-less regimes (as well as averages in between) are all dictatorships in a dichotomonous transformation. With respect to validity, the alternative measure for regime-type is not necessarily a better proxy for “democracy” in general. As Jaggers and Gurr (1995) clearly suggest, different measures of regime-type are effective for different research purposes. If the only objective is to classify regimes, a true measure should include institutions and outcomes, both of which can be useful. For example, the effective party/constitutional framework measure is probably not as effective as the Gastil scale at measuring human rights abuses and civil rights, a distinct scholarly endeavor from explaining growth. “Democracy” is a slippery concept with several meanings.14 Given this particular econometric question, the effective party/constitutional framework measure is useful in two aspects. First, continuous expression of the proxy not only maintains its intuitive appeal, unlike the outcome measures, but also makes key distinctions where the Gastil and Polity III measures do not. For example, with respect to democracies, the rankings in Table 1 show considerable variation in the effective party/constitutional framework measure among the most advanced pluralistic democracies, while the Gastil and Polity III produce identical ordinal values for fourteen (“1”) and twenty-two (“10”) countries, respectively. The outcome measures assign a considerable population of democratic countries with indistinguishable values, while the alternative institutional measure quantitatively differentiates party systems and presidential and parliamentary democracies.15 In addition to the continuous differentiation of democracies, to distinguish between authoritarian regimes is an important objective because econometric studies exclusively consider it a residual category. For example, Barro writes that “the theory that determines which kind of dictatorship will prevail is missing” (1996, p. 2). Similarly, Przeworski and Limongi cite Sah’s contention that economic performance varies more dramatically among dictatorships 88 DURHAM Table 1. Regime-type measures: Effective party/constitutional framework, Gastil, and Polity III: five-year average, 1960–1989.a Country Belgium Denmark Netherlands Papua New Guinea Israel Iceland Italy Norway Sweden Finland West Germany Japan Switzerland Ireland Australia Thailand Sir Lanka Turkey Canada Austria France Mauritius United Kingdom India Cyprus Greece New Zealand Venezuela Argentina Costa Rica Guatemala Colombia South Korea Jamaica Ecuador Bolivia Trinidad and Tobago Nicaragua El Salvador United States Brazil Gambia Philippines Dominican Republic Panama Effective Party/ Constitutional Framework Measure 4.885 4.734 4.540 4.521 4.133 3.805 3.530 3.347 3.254 3.085 2.997 2.712 2.681 2.609 2.491 2.396 2.372 2.356 2.314 2.258 2.243 2.174 2.105 2.080 2.076 2.025 1.970 1.930 1.738 1.714 1.564 1.533 1.527 1.513 1.497 1.492 1.488 1.481 1.468 1.462 1.447 1.441 1.436 1.417 1.409 a. The Gastil measures begin in 1973. (Rank) Gastil Ordinal Measure 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 1.000 1.000 1.000 2.000 2.125 1.000 1.375 1.000 1.000 1.750 1.250 1.250 1.000 1.125 1.000 3.000 3.500 2.750 1.000 1.000 1.500 2.375 1.000 2.375 2.125 1.667 1.000 1.750 1.500 1.000 3.000 2.625 5.167 2.125 3.000 3.500 1.750 4.250 3.000 1.000 3.500 2.500 4.000 2.167 5.375 (Rank) Polity III Ordinal Measure (Rank) 1 1 1 26 27 1 18 1 1 23 16 16 1 15 1 38 42 37 1 1 19 31 1 31 27 22 1 23 19 1 38 35 72 27 38 42 23 55 38 1 42 33 51 30 74 10.000 10.000 10.000 10.000 9.233 10.000 10.000 10.000 10.000 10.000 10.000 10.000 10.000 10.000 10.000 −0.600 5.760 8.667 10.000 10.000 7.560 9.400 10.000 8.433 9.400 7.400 10.000 8.400 3.500 10.000 −1.650 7.533 −5.900 10.000 2.680 −1.450 8.300 −6.250 0.900 10.000 −3.000 9.400 0.733 0.400 −3.600 1 1 1 1 26 1 1 1 1 1 1 1 1 1 1 47 35 27 1 1 32 23 1 28 23 34 1 29 38 1 52 33 64 1 39 51 30 70 41 1 54 23 42 44 57 89 ECONOMIC GROWTH AND POLITICAL REGIMES Table 1. Continued. Country Peru Paraguay Zimbabwe Botswana Guyana Malaysia Bangladesh South Africa Uruguay Chile Sierra Leone Egypt Indonesia Mexico Portugal Taiwan Zambia Poland Senegal Honduras Singapore Kenya Hungary Liberia Yugoslavia USSR Czechoslovakia Bulgaria Sudan Togo China Romania Mozambique Cameroon Congo Tunisia Algeria East Germany Malawi Zaire Spain Pakistan Syria Iraq Benin Myanmar Mali Ghana Haiti Effective Party/ Constitutional Framework Measure 1.398 1.384 1.377 1.369 1.367 1.351 1.346 1.340 1.310 1.300 1.225 1.196 1.150 1.145 1.128 1.126 1.087 1.079 1.071 1.030 1.005 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.989 0.925 0.879 0.755 0.750 0.750 0.667 0.527 0.502 (Rank) Gastil Ordinal Measure 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 67 67 67 67 67 67 67 67 67 67 67 67 67 67 67 67 67 67 86 87 88 89 90 90 92 93 94 3.833 4.625 4.750 2.500 3.750 3.833 4.000 5.500 3.750 4.000 5.000 4.500 5.125 3.750 1.500 4.625 5.125 4.833 4.125 3.500 4.625 5.250 4.833 5.500 5.375 6.250 6.500 7.000 5.500 6.125 6.333 6.667 6.500 5.750 6.500 5.000 5.750 6.500 6.375 6.500 2.667 4.625 6.333 6.833 6.750 6.833 6.333 6.000 6.500 (Rank) Polity III Ordinal Measure (Rank) 49 57 61 33 46 49 51 76 46 51 65 56 69 46 19 57 69 62 54 42 57 73 62 76 74 87 92 104 76 86 88 99 92 82 92 65 82 92 91 92 36 57 88 101 100 101 88 84 92 2.440 −7.800 0.100 10.000 −3.100 7.867 −5.400 4.000 4.350 −0.900 −4.680 −5.400 −6.880 −4.800 0.400 −7.400 −6.240 −6.067 −3.360 0.500 −2.000 −5.520 −6.200 −6.200 −6.200 −6.520 −6.950 −7.000 −7.000 −7.000 −7.133 −7.400 −7.600 −7.640 −8.000 −8.240 −8.600 −9.000 −9.000 −9.000 −0.720 −0.833 −8.500 −7.100 −6.500 −7.250 −7.000 −5.700 −9.200 40 90 46 1 55 31 60 37 36 50 58 60 73 59 44 85 69 65 56 43 53 62 66 66 66 72 74 75 75 75 83 85 87 89 92 94 97 98 98 98 48 49 96 82 71 84 75 63 101 90 DURHAM Table 1. Continued. Country Rwanda Niger Uganda Nepal Lesotho Iran Central African Republic Jordan Kuwait Bahrain Swaziland Effective Party/ Constitutional Framework Measure 0.500 0.333 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (Rank) Gastil Ordinal Measure 95 96 97 97 97 97 97 97 97 97 97 6.000 6.500 7.000 4.833 5.125 5.500 6.833 5.500 5.000 5.000 5.500 (Rank) Polity III Ordinal Measure (Rank) 84 92 104 62 69 76 101 76 65 65 76 −7.000 −7.000 −7.000 −7.600 −7.950 −8.100 −8.333 −9.300 −9.600 −10.000 −10.000 75 75 75 87 91 93 95 102 103 104 104 than among democracies, largely because some dictators are constructively “developmentalist,” while others are simply “thieves.” Similar to Barro (1996) and Alesina et al. (1996), they lament that “we have no theory that would tell us in advance which we are going to get” (Przeworski and Limongi, 1993, p. 65). By quantitatively distinguishing authoritarian regimes, this measurement begins to address the lack of testable hypotheses. The effective party distinction does not purport to truly distinguish benevolent technocratic authoritarian regimes from malevolent kleptocratic despots. But given its crucial link to executive discretion—party-less regimes enjoy more discretion than single-party dictatorships—the proxy at least provides an objective hypothesis on which type should outperform, depending on whether one generally espouses the authoritarian or democratic view of development. If discretion is benevolent, then party-less dictatorships should grow faster, otherwise, checks on arbitrary capriciousness in single-party authoritarian regimes should lead to better performance. In short, the effective party/constitutional framework measure makes a first cut at objective econometric distinctions and therefore testable hypotheses regarding subtypes of both democracies and dictatorships. A second advantage regards its more explicit institutional focus. While outcome measures address a broader conception of democracy and a fuller spectrum of freedoms, an institutional proxy focuses more acutely on regime type characteristics that relate to the origins of government policymaking and constitutional framing. Whether democracies or dictatorships grow faster is not a purely academic inquiry. The effective party/constitutional framework measure complements common outcome proxies because the former more directly addresses the question, perhaps initially motivated by North and Thomas (1973), as to whether social scientists can identify objectively measurable institutions that promote growth. Policy entrepreneurs cannot plausibly construct a polity to instantaneously possess some value on an ordinal outcome scale. They cannot directly compel such results, but they can craft procedural institutions.16 With respect to the political parties, the first component, the alternative measure ultimately relates to the constitutional choice in voting procedures ECONOMIC GROWTH AND POLITICAL REGIMES 91 among first-past-the-post, proportional representation and various combinations thereof. Also, some dictators have the option to surround themselves with political party machinery. Furthermore, with respect to the second component, constitutional frameworks ultimately refer to serviceable stipulations—actual choices regarding the relationship between the executive and legislative government functions. In short, economists and policy entrepreneurs alike can supplement outcome proxies with institutional measures that have more concrete empirical and applicable referents. 2.4. Hypotheses: Linear and Nonlinear Forms The continuous nature of the effective party/constitutional framework measure enables tests of three hypotheses.17 First, the relationship between the number of effective parties and growth might be linear, which perhaps becomes less intuitive after certain values in the independent variable. One might question whether there is a monotonic effect of parties on growth after some threshold. Put differently with respect to specific examples, the (beneficial or deleterious) effect of the 4.1 effective parties in Israel compared to the 2.1 in the United Kingdom might be less pronounced than the approximately equivalent difference between the United Kingdom and a party-less regime such as Jordan. Thus, possibility of diminishing marginal positive or negative effects recommends investigating curvilinear functional forms. Preserving the direction of a given hypothesis, linear-log forms test whether the effect on economic growth becomes less pronounced or tails off after some democratic (or discretion) threshold. This form suggests that the most pronounced effect occurs form zero to one effective party, which might not as accurately reflect this nonlinearity. In contrast to the linear and linear-log forms, the effect of political regimes might not be unidirectional. Another curvilinear form evaluates a more balanced view. A concave quadratic form tests whether the effect on economic growth is positive at an increasing rate approaching a critical value of effective parties and then tapers increasingly downward. More substantively, the quadratic form suggests that governing institutions must restrain sovereigns, compel a credible commitment to property rights, but also restrain them from wasteful consumption. A certain level of party development or competition ensures commitment, but too many parties make the state vulnerable to demands for consumption.18 One might expect that in terms of the effective party/constitutional framework measure, the critical maximum might be between one and two parties, which should reflect the existence of a credible opposition but yet sufficient discretion for unpopular (and presumably ultimately benevolent) measures. Less intuitively but still quite possibly, a critical maximum between, say, three or four parties would suggest that democracies generally outperform dictatorships, with the exception of the most fragmented systems. 3. Additional Econometric Issues: Developmental “Stages” and Simultaneity and/or Survivor Bias Before econometric analysis of the effective party/constitutional framework proxy, two crucial issues that studies fail to address must be outlined. The first regards the possibility that the efficacy of particular regimes is circumstantial rather than universal, as no study tests whether dictatorships (democracies) are more efficient in less (more) developed areas. The 92 DURHAM second concerns the misguided critique that studies (Kormendi and Meguire, 1985; Grier and Tullock, 1989; Barro, 1996) are prone to simultaneity and/or survivor bias. Contrary to this sweeping criticism (Brunetti, 1997; Przeworski and Limongi, 1993, 1997), given the routine distinction between the level and rate of economic development, regime type is exogenous to growth. 3.1. The “Stages” of Economic Development The literature fails to test persuasive but implicit notions about the “stages” of development. Again, a given sovereign must be strong enough to protect property rights and prevent the anarchical state of nature, yet institutions must also restrain them from abusing monopoly power for unproductive ends. These paradoxical requirements are important, yet debaters typically emphasize only one component of this relationship between sovereign and subject and make universal assertions. But both characteristics might seemingly have organizational advantages under certain circumstances,19 and unfortunately growth econometrics render no explicit evaluations of possible conditions under which discretion is beneficial. Drawing on sweeping generalizations of economic history, an historical growth trajectory illustrates more implicit hypotheses about the conditions under which some regimes are optimal. Consider Kurth’s (1979) outline of industrial phrases and the concomitant political outcomes. These successive phases include “the production of non-durable consumer goods, intermediate and capital goods, and consumer durables” (Kurth, 1979, p. 319). Briefly, the first and last stages are complementary for democracy because investment requires “relatively modest” sums of capital in relation both to the available stock and the large amount required for industrial development (Kurth, 1979, p. 330). Some scholars, then, Gerschenkron certainly included, instinctively associate the greater amount of capital required and attendant extensive state intervention during the second phase with authoritarianism. On the one hand, most agree that today the wealthiest countries are democracies (Przeworski and Limongi, 1993, p. 62). Perhaps consequently, few scholars explicitly argue that advanced democracies, such as the United States or Japan, would grow faster if they were authoritarian.20 This largely conforms to Kurth’s suggestion that democracies are more efficacious in the latest phase. Conversely, even in the paradigmatic case of liberal nineteenth-century Great Britain, few economic historians argue that the process of heavy industrial development occurred under truly democratic auspices. Granted, this assertion depends on one of many possible operationalizations of democracy, but if procedural universal suffrage and full working-class inclusion are indices, then the generalization largely holds. For example, Schwarz (1992) suggests that OECD countries industrialized under less than liberal political tutelage, while democratization followed heavy industrial development. Even Moore (1966), who differentiates democratic, fascist, and communist development, notes that industrialization proceeded universally without subject approval. He writes, “There is no evidence that the mass of the population anywhere has wanted an industrial society. . . . At bottom of all forms of industrialization so far have been revolutions ECONOMIC GROWTH AND POLITICAL REGIMES 93 from above, the work of a ruthless minority” (p. 506). This hardly suggests a democratic option in the early phases. On the other hand, few economic historians counterfactually argue that heavy industrial development would have been more efficient under democratic auspices. Even Kurth’s sweeping trajectory aside, the supposed virtues of democracy seem more relevant in comparatively advanced stages of development. The democratic view largely assumes a modicum of wealth, existing entrepreneurs, and readily available capital. In more destitute countries, there are simply fewer enterpreneurs and households with the wherewithal to invest, and the credible commitment problem is perhaps comparatively less pressing. In poor countries, one might argue that the shear dearth of capital supersedes the commitment to credible property rights. After all, the establishment of property rights generally coincides with the development of property (capital) itself. Put differently, the ratio of unskilled labor to the total human capital stock decreases with development (national income), which implies that dictatorships, given their purported advantage in compelling savings from underprivileged classes, become increasingly ineffective and costly. This review does not suggest that the literature explicitly conforms to Kurth’s outline, but neither do scholars evaluate the tacit contention that dictatorship is more efficacious during heavy industrial development or that democracies better govern advanced states. Critics and defenders of each regime type talk past one another if one considers (very) long-term growth and development. Therefore, one might expect that in the universal sample, the effect of regime type should be ambiguous because the negative effect of discretion in advanced countries vitiates the positive impact in poorer areas. Such consensus, however implicit, invites explicit evaluation. 3.2. Simultaneity Bias, Survival Bias, or Collinearity? Just as many social scientists argue that political regime type affects economic development, many others also advance the notion that economic development affects political regime type. Similar to Brunetti’s (1997) discussion, Przeworski and Limongi’s (1993) conclusion is that the evidence of simultaneous determination and survival bias between political regime type and development “suffices to render suspect any study that does not treat regimes as endogenous.”21 This profound criticism applies to Kormendi and Mequire (1985), Grier and Tullock (1989), and Barro (1996). But briefly, an important yet routine distinction between the level and rate of economic development shows that the estimation problem involves neither simultaneity nor survival bias but collinearity between the initial level of income and regime type.22 Again, the debate concerns whether regime type determines growth. That is, either dictatorships or democracies exhibit better economic performance. To suggest that economic development in turn affects regime type necessarily implies that development determines the manifestation of either democracy or dictatorship. Economic development determines a particular regime type. Brunetti (1997) as well as Przeworksi and Limongi (1993, 1997) do not satisfactorily evaluate the distinct (purported) effects of the rate and level of economic development. Whether the rate, level, or both aspects of economic development help determine political regime type implies different econometric problems and solutions. If 94 DURHAM the level of development determines regime type, a regression that explains growth as some exogenous function of regime type should control for the initial level of development, which in turn should exhibit some degree of collinearity with regime type. If growth is jointly determined with regime type, as the growth rate must help determine regime type not regime longevity, then the statistical problem is simultaneity bias. That is, growth must specifically affect democratic or authoritarian outcomes, not the general likelihood of regime survival. While the econometric section explores this question anew, a brief review of published results is instructive. Scholars thoroughly document the empirical relationship between the level of economic development and democracy.23 Theory is less complete,24 but enough evidence warrants evaluating the relationship between the level of development and regime type because the vast majority of developed nations are democratic under most definitions. Levels of development correlate with instances of democracy, a particular regime type. But no published evidence suggests that the rate of economic development affects regime type.25 Growth rates do not correlate with manifestations of democracy or dictatorships.26 Perhaps more important, Przeworski et al. (1995) as well as Przeworski and Limongi (1997) provide no theoretical reason why democracies should be more sensitive than dictatorships to poor economic performance in the short–run but not to mention the long run. Contrarily, Bardhan suggests that authoritarian regimes are actually less stable under deteriorating performance because they lack popular legitimacy. He notes the possibility that authoritarian regimes might survive during periods of sustained economic growth, but when an economy persistently underperforms, democracies sometimes survive because they possess popular legitimacy, whereas dictators do not (Bardhan, 1994, p. 237). Empirical referents might include the “bureaucratic-authoritarian” model in Latin America or Marx’s explanation for Bonapartism in nineteenth-century France, instances in which economic performance is the regime’s raison d’être. 4. Econometric Results Given the three theoretical addenda, the final issue of possible simultaneity bias in these data must be addressed before considering the direct effect of regimes on growth and investment in the total sample, across income breaks to address development stages, and exclusively within single-party and party-less dictatorships. Each model employs a GLS random-effects procedure,27 and to ameliorate some validity issues associated with such (unbalanced) “panel” models the regressions follow Grier and Tullock’s (1989) design with five-year averaging for the medium-run specifications and dummy variables for continent and time period to absorb specific spatial and temporal effects, respectively.28 4.1. Regime Type Measures Regressions on each democracy proxy, which include many of the factors that Barro (1996), Przeworski et al. (1995), and Helliwell (1992) employ, statistically support the conclusion regarding the exogenous nature of political regime type. More formally, the purported ECONOMIC GROWTH AND POLITICAL REGIMES 95 simultaneous equation system includes Y1 = β0 + β2 Y2 + β3 Y3 + β4 X + µ (3) Y2 = α0 + α1 Y1 + α3 Y3 + α4 Z + ε, (4) and where Y1 is economic growth, Y2 is some measure for political regime type, Y3 is the initial level of development, X refers to the set of variables that identifies the first equation, and Z is the set of variables that identifies the second equation. To determine whether a simultaneous procedure is empirically necessary to estimate Y1 , α1 should be statistically significant. If not and especially given no persuasive theory as discussed in the previous section, little recommends substituting the instrumental variables derived in (4) for the actual regime type measure in (3). If α3 is significant, the coefficient for the initial level of development, then Y2 and Y3 , the measures of regime type and the initial level of development, respectively, likely exhibit some degree of collinearity in growth regressions such as (3). Regressors include the growth rate, the initial level of economic development, male and female education levels, and factors that identify (4) with respect to (3)—the proportions of contemporary democracies in the continent of a given country and the rest of the world, a dummy for British colonial heritage and, in some specifications, lagged values of Y2 .29 The regressions30 include short-, medium-, and long-run averages, which respectively imply two panels and a pure cross-sectional design. Turning to the results, the coefficient of the growth rate α1 is not statistically significant in any regression for regime type Y2 , while the coefficient for the log initial level of development α3 is significant within the 10 percent level in every pooled but in no cross-sectional regression. This includes alternative regressions on the Polity III data, the number of effective political parties, and logistic regressions on the Polity III dummy variable in a pure cross-sectional mode.31 Perhaps more persuasively, this analysis considers a subset that solely includes the poorest areas, which addresses the contention in Przeworski et al. (1995) that more destitute democracies are particularly fragile. But no regression for countries with per capita incomes of less than $2,000 or $1,000 yields a statistically significant coefficient for the growth rate α1 , while the coefficient for the log initial level of development α3 is significant in all but one regression.32 Again, these include short-, medium-, and long-run models; alternative considerations of Polity III and effective party data. In short, given comparatively consistent significant values for α3 , the results clearly indicate that the level and not the rate of development correlates with regime type. Specifically, to confirm Lipset’s (1959) hypothesis and Barro’s (1996) more recent findings, both of which do not control for growth, more developed nations are more likely to exhibit democratic characteristics. This further recommends inclusion of initial per capita incomes in the neoclassical equations on growth and suggests the possibility of some collinearity, but a simultaneous equation procedure is neither theoretically essential nor empirically necessary, which contradicts Brunetti (1997) and Przeworski and Limongi (1993, 1997).33 Political regime type is therefore exogenous to growth. 96 4.2. DURHAM Growth and Investment: Total Sample Given no simultaneity bias, the growth and investment equations extend a neoclassical34 model to include regime type.35 As Levine and Renelt (1992) suggest, with approximately fifty statistically significant growth regressors reported in the literature, specification bias is prevalent (p. 943). Indeed, even among carefully specified studies that address this question such as Kormendi and Meguire (1985), Grier and Tullock (1989), and Barro (1996), there is no precisely common set of controls. This study follows Levine and Renelt’s four “base” regressors that economists most commonly include in growth regressions, including the log initial level of development, the level of investment in physical capital, the level of investment in human capital (male and female education rates),36 and population growth. However sensitive or empirically insignificant, two more controversial macroeconomic variables control for relevant theoretical considerations particularly germane to executive discretion, including the government ratio and openness to world trade.37 First, “supplyside” theories argue that increased taxes and government spending distort incentives and reduce the efficiency of capital allocation. Regardless, any government more involved in the economy, irrespective of regime type, would more likely have an effect on macroeconomic performance generally and growth specifically, however virtuous. The model must distinguish regime type and the general effects of government intervention.38 Second, classical liberal economic theorists such as Ricardo advance the virtues of comparative advantage. More recently, Romer expounds this view with a different emphasis. In contrast to the conventional focus on “objects,” Romer suggests that international trade creates greater prosperity through the diffusion of productive ideas into the periphery (1993, p. 546). But besides the economic rationale, this variable addresses the impact of international influences on policy choice—exposure to the world economy39 constrains sovereigns and thereby limits their discretion (Gourevitch, 1978).40 Turning to the results, as the total sample regressions in Table 2 indicate, none of the three alternative functional forms of the effective party/constitutional framework measure are statistically significant—which generally differs from Kormendi and Meguire (1985), Grier and Tullock (1989), and Barro (1996). The linear and linear-log forms indicate positive yet insignificant relationships, and unlike Barro’s finding (1996, p. 14), the quadratic form is also insignificant and suggests no optimal degree of discretion. Besides inferences regarding the regime type proxy, the three equations substantiate enduring neoclassical hypotheses. That is, the evidence generally supports the convergence hypothesis, the benefits of investment and openness to international trade, and the negative effects of population growth and government spending. Also, the proxy for male education rates is positive, and female education rate is negative, consistent with previous findings. But both human capital measures are insignificant. As Levine and Renelt (1992) note, scholars commonly advance certain growth determinants that relate more intuitively to investment. Similarly, the argument that dictatorships grow faster than democracies suggests that authoritarian regimes avoid electoral cycle consumption, which in turn facilitates increases in the physical capital stock. Conversely, the democratic view argues that entrepreneurs have the incentive to invest under secure property rights. Like the growth equations in Table 2, regressions listed in Table 3 suggest 97 ECONOMIC GROWTH AND POLITICAL REGIMES Table 2. Growthb regressions: Total sample, GLS random effects models: Medium-run (five-year averages), 1960–1989. Regression 1 β Independent Variable Effective party/constitutional framework (EP/CF) Log EP/CF Square EP/CF 0.043 p value Regression 2 β p value 0.774 0.009 Regression 3 β p value 0.078 0.803 −0.008 0.897 0.828 Log initial GDP Investment ratio Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept −1.077 0.110 0.173 −0.077 −0.443 0.008 −0.093 −0.814 0.000 0.000 0.383 0.708 0.010 0.042 0.000 0.130 −1.053 0.109 0.170 −0.074 −0.442 0.008 −0.092 −0.835 0.000 0.000 0.401 0.721 0.010 0.040 0.000 0.122 −1.070 0.110 0.172 −0.078 −0.442 0.008 −0.093 −0.821 0.000 0.000 0.386 0.707 0.010 0.043 0.000 0.127 −1.245 −1.797 −1.273 −2.001 10.278 0.047 0.004 0.041 0.061 0.000 −1.273 −1.810 −1.262 −1.991 10.173 0.045 0.004 0.040 0.060 0.000 −1.259 −1.806 −1.275 −2.008 10.210 0.046 0.004 0.040 0.059 0.000 Observations Cases Mean time periods per case 450 105 4.286 450 105 4.286 450 105 4.286 R2 within R2 between R2 overall 0.359 0.435 0.417 0.359 0.436 0.417 0.359 0.437 0.417 b. The source for all growth calculations is the inital GDP values from the Summers and Heston data set. that neither the linear, linear-log, nor quadratic forms of the effective party/constitutional framework measure are statistically significant in the total sample.41 4.3. Developmental “Stages” Another econometric objective is to explicitly test hypotheses regarding the levels of development. Indeed, this distinction perhaps directly addresses the dearth of significant findings in the complete sample. That is, given that political regime type (discretion) does not apparently affect growth or investment universally, the organizational efficiency of dictatorship or democracy possibly depends on certain circumstances. The malevolent effect of discretion in rich countries perhaps neutralizes its benevolent impact in more destitute cases, which produces the ambiguous universal effect reported in Section 4.2. “Phases” or “stages” of industrialization are slippery concepts, but this inductive exercise 98 DURHAM Table 3. Investmentc regressions: Total sample, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Regression 4 Independent Variable β p value Regression 5 β p value Regression 6 β EP/CF Log EP/CF Square EP/CF −0.326 Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 1.936 0.694 0.211 0.736 0.056 −0.096 −4.283 0.009 0.110 0.651 0.019 0.000 0.048 0.005 1.674 0.557 0.328 0.751 0.057 −0.101 −4.519 −3.836 −0.770 5.797 1.304 −5.917 0.032 0.677 0.001 0.679 0.293 −4.227 −1.045 5.355 0.595 −3.771 Observations Cases Mean time periods per case 451 105 4.295 451 105 4.295 451 105 4.295 R2 within R2 between R2 overall 0.164 0.725 0.622 0.166 0.725 0.622 0.170 0.732 0.630 0.294 0.117 p value 0.753 0.197 −0.242 0.029 0.021 0.204 0.482 0.017 0.000 0.035 0.003 1.844 0.644 0.232 0.747 0.059 −0.097 −4.494 0.012 0.137 0.617 0.017 0.000 0.043 0.003 0.018 0.571 0.002 0.848 0.501 −4.135 −0.972 5.877 1.396 −5.843 0.020 0.595 0.001 0.653 0.295 0.153 c. The source for all investment data is from the Summers and Heston data set. is to divide the sample across eight percentiles of initial per capita GDP levels.42 That is, “less developed” samples include countries below the twentieth ($1,114), thirtieth ($1,487), fortieth ($2,093), and fiftieth ($2,904) percentiles. Conversely, the “developed’ samples include countries above the eightieth ($7,638), seventieth ($5,463), sixtieth ($3,892), and fiftieth ($2,904) percentiles. This division of the sample hardly captures subtleties such as “industrial profiles” or sufficient “backward linkages,” but nonetheless countries above (below) certain income breaks should be relatively more (less) developed, however complex the concept. Again, the hypothesis is that discretion is effective in less developed areas but deleterious in advanced economies with respect to both the allocation and accumulation of resources.43 The data strongly contradict the view that discretion is more suitable in less developed areas. According to Table 4 and Table 5, neither linear nor linear-log forms of the effective party/constitutional framework measure is significant in growth regressions using any less developed sample break. More problematic for Kurth and the tacit consensus, executive discretion actually inhibits capital accumulation in poorer areas. For example, the linear- 99 ECONOMIC GROWTH AND POLITICAL REGIMES Table 4. Growth regressions: Less developed sample (20%, 30%, 40%, and 50% per capita GDP percentiles), Linear effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Per Capita GDP Level: Regression 7 <$1,114 Regression 8 <$1,487 β Regression 9 <$2,093 Independent Variable β p value EP/CF Log initial GDP Investment ratio Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept −0.223 −0.589 0.190 −0.724 1.588 −0.254 −0.013 −0.040 −1.165 0.662 0.596 0.000 0.083 0.013 0.635 0.327 0.309 0.172 −0.473 0.496 0.172 −0.493 0.655 −0.066 −0.008 −0.085 −1.514 0.196 0.541 0.000 0.122 0.150 0.803 0.468 0.011 0.038 −0.302 0.164 0.144 0.022 0.025 0.007 0.005 −0.086 −1.198 0.390 0.814 0.000 0.945 0.954 0.980 0.549 0.012 0.146 −0.390 0.580 0.131 −0.041 0.047 0.099 0.006 −0.089 −1.588 0.129 0.181 0.000 0.857 0.866 0.650 0.386 0.000 0.004 −2.641 0.083 −2.772 0.009 −1.981 0.057 −2.639 0.000 −3.900 0.088 −2.537 −1.559 0.025 0.380 6.198 0.382 0.722 0.888 −2.659 −1.422 −2.074 0.881 0.012 0.365 0.437 0.856 −2.889 −1.296 −2.097 −1.660 0.000 0.182 0.307 0.584 p value β Regression 10 <$2,904 p value β Observations Cases Mean time periods per case 88 31 134 46 179 59 224 70 2.839 2.913 3.034 3.200 R2 within R2 between R2 overall 0.304 0.881 0.591 0.268 0.586 0.500 0.308 0.445 0.431 0.308 0.638 0.471 p value log form equations for the twentieth, thirtieth, and fiftieth percentiles (Regressions 19, 20, and 22 in Table 7) suggest that increases in the number of effective parties increases the investment ratio at a decreasing rate in less developed countries. Furthermore, while the linear form is insignificant in the more affluent remaining breaks, the twentieth percentile (under $1,114) equation, Regression 15 in Table 6, indicates a significantly positive linear effect. In short, however robust these unexpected results, the data clearly reject the view that dictatorship is more suitable for poorer economies. With respect to advanced countries, the data render some support for the trajectory hypothesis. For example, the linear-log form is consistently significant in growth regressions for the three highest percentiles (Regressions 27, 28, and 29 in Table 9), which suggests a positive impact on expansion in more advanced economies. Less compelling, the linear form is also positive and significant, albeit within the 10 percent confidence interval, in samples above the seventieth and sixtieth income percentiles (Regressions 24 and 25). But interestingly, discretion only seems to inhibit resource allocation but not accumulation in advanced areas. The coefficients for the linear and linear-log forms are insignificant in 100 DURHAM Table 5. Growth regressions: Less developed sample (20%, 30%, 40%, and 50% per capita GDP percentiles), Linear-log effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (fiveyear averages), 1960 to 1989. Per Capita GDP Level: Regression 11 <$1,114 Regression 12 <$1,487 β Regression 13 <$2,093 Independent Variable β p value EP/CF Log initial GDP Investment ratio Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 0.021 −0.681 0.178 −0.830 1.763 −0.272 −0.012 −0.036 −1.184 0.784 0.537 0.001 0.048 0.006 0.612 0.340 0.361 0.170 −0.037 0.313 0.175 −0.529 0.664 −0.112 −0.009 −0.085 −1.344 0.560 0.693 0.000 0.104 0.158 0.671 0.424 0.013 0.076 −0.027 0.311 0.146 0.028 −0.005 0.122 0.007 −0.077 −1.415 0.662 0.649 0.000 0.930 0.990 0.636 0.409 0.022 0.085 −0.056 0.502 0.134 −0.018 0.011 0.086 0.006 −0.090 −1.435 0.266 0.246 0.000 0.938 0.968 0.695 0.379 0.000 0.011 −2.721 0.077 −2.589 0.018 −2.176 0.035 −2.509 0.000 −3.954 0.085 −2.578 −1.333 0.027 0.463 6.735 0.343 1.468 0.783 −2.849 −1.148 −3.395 −0.958 0.006 0.435 0.145 0.842 −2.829 −1.242 −3.171 −1.713 0.000 0.203 0.090 0.578 p value β Regression 14 <$2,904 p value β Observations Cases Mean time periods per case 88 31 134 46 181 60 224 70 2.839 2.913 3.017 3.200 R2 within R2 between R2 overall 0.304 0.886 0.591 0.269 0.574 0.493 0.293 0.488 0.435 0.308 0.632 0.468 p value every investment equation for developed samples, with the exception of the linear form in the third highest percentile (above $3,892, Regression 33), which is unexpectedly negative. 4.4. Growth and Investment in Dictatorships Again, the preceeding discussion suggests that the effective party/constitutional framework measure quantitatively distinguishes different authoritarian regimes. To extend Jowitt’s and Huntington’s arguments about corruptive tendencies in political parties, the testable hypothesis, under the general authoritarian perspective, is that party-less dictatorships promote growth and investment more effectively than single-party regimes. Conversely, according to the broad democratic view, single-party regimes should be more effective than party-less despots. Turning to the results, with respect to allocation, Regression 39 in Table 12 suggests that single-party regimes had slightly higher growth rates, but the figure is insignificant. But regarding capital accumulation, single-party regimes had investment ratios on average 101 ECONOMIC GROWTH AND POLITICAL REGIMES Table 6. Investment regressions: Less developed sample (20%, 30%, 40%, and 50% per capita GDP percentiles), Linear effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Per Capita GDP Level: Regression 15 <$1,114 Regression 16 <$1,487 β Regression 17 <$2,093 Independent Variable β p value EP/CF Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 2.344 2.750 0.298 −0.716 −2.510 0.195 −0.050 −7.301 0.011 0.260 0.795 0.710 0.022 0.000 0.535 0.011 1.293 2.226 −0.430 0.269 −0.362 0.124 −0.080 −4.711 0.129 0.288 0.608 0.835 0.485 0.000 0.335 0.066 0.105 3.188 −0.156 1.836 −0.112 0.068 −0.072 −1.740 0.869 0.026 0.814 0.037 0.807 0.000 0.274 0.365 0.452 2.934 0.085 1.085 0.006 0.062 −0.115 −2.633 0.399 0.011 0.887 0.132 0.989 0.000 0.048 0.110 −8.516 0.075 −5.149 0.166 −4.265 0.080 −3.553 0.089 13.341 0.068 0.860 17.558 0.826 0.009 −7.725 0.620 −6.864 0.615 0.050 8.597 0.400 −15.078 0.984 0.012 0.946 0.134 0.302 7.890 −0.837 −12.194 0.889 0.002 0.877 0.129 p value β Regression 18 <$2,904 p value β Observations Cases Mean time periods per case 88 31 134 46 179 59 224 70 2.839 2.913 3.034 3.200 R2 within R2 between R2 overall 0.617 0.469 0.414 0.321 0.396 0.349 0.257 0.548 0.414 0.221 0.645 0.479 p value and ceteris paribus about three percentage points greater than their party-less counterparts during the period, which is significant at the 5 percent level according to Regression 40. This contradicts Jowitt’s and Huntington’s views based on more qualitative evidence, and the results suggest, in support of the general democratic view, that greater discretion within dictatorships enhances investment.44 A plausible interpretation is that institutionalized parties on average restrain capricious autocrats and thereby create a more suitable investment environment. 5. Conclusions In summary, a crucial shortcoming in the debate on regimes and growth concerns the dearth of continuous measures that focus on institutions rather than outcomes. While not necessarily a superior proxy for all econometric applications, the institutional focus of the effective party/constitutional framework measure more directly addresses the serviceable implications of this particular debate. Also, two additional problems beset previous studies. First, the literature ignores the possibility that the effectiveness of particular regime types 102 DURHAM Table 7. Investment regressions: Less developed sample (20%, 30%, 40%, and 50% per capita GDP percentiles), Linear-log effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (fiveyear averages), 1960 to 1989. Per Capita GDP Level: Regression 19 <$1,114 Regression 20 <$1,487 β Regression 21 <$2,093 Independent Variable β p value EP/CF Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 0.287 2.884 0.448 −0.817 −2.237 0.195 −0.048 −7.780 0.022 0.245 0.697 0.671 0.045 0.000 0.562 0.006 0.263 2.001 −0.408 0.222 0.295 0.131 −0.074 −5.487 0.037 0.333 0.623 0.862 0.563 0.000 0.368 0.031 0.150 2.981 −0.268 1.857 −0.115 0.069 −0.068 −2.156 0.187 0.036 0.685 0.033 0.801 0.000 0.297 0.264 0.198 2.860 −0.030 1.124 0.028 0.064 −0.109 −3.054 0.065 0.012 0.960 0.116 0.948 0.000 0.058 0.064 −8.870 0.059 −5.648 0.123 −4.570 0.058 −3.928 0.059 12.687 0.078 0.858 16.875 0.824 0.011 −6.886 0.668 −3.896 0.776 −0.046 8.827 −0.065 −12.926 0.985 0.010 0.990 0.201 0.247 8.006 −0.312 −10.663 0.908 0.002 0.951 0.184 p value β Regression 22 <$2,904 p value β Observations Cases Mean time periods per case 88 31 134 46 179 59 224 70 2.839 2.913 3.034 3.200 R2 within R2 between R2 overall 0.602 0.474 0.419 0.324 0.417 0.371 0.261 0.558 0.422 0.226 0.654 0.488 p value depends on certain circumstances—namely, the level of economic development, which perhaps helps explain ambiguous findings in universal samples. Second, these data contradict the notion that regimes and growth are jointly determined (Brunetti, 1997; Przeworski and Limongi, 1993, 1997), as the level rather than the rate of development affects regimes. Therefore, Kormendi and Meguire (1985), Grier and Tullock (1989), Barro (1996), and this article correctly consider regime type exogenous to growth. Given these addenda, the total sample renders no empirical regularities between any functional form of the number of effective political parties and growth or investment in the total sample. But this does not necessarily suggest that political regime type does not affect performance. Division of the sample across eight per capita income breaks to inductively proxy “stages” of development produces some significant results. As expected, the linearlog form suggests that discretion negatively affects growth in developed countries. But in direct contrast to Kurth’s hypothesis, some evidence indicates that discretion actually lowers investment in poorer areas. Also, given substantively meaningful authoritarian subtypes, hitherto an econometric residual category, single-party dictatorships have greater investment ratios compared to party-less regimes. 103 ECONOMIC GROWTH AND POLITICAL REGIMES Table 8. Growth regressions: Developed sample (80%, 70%, 60%, and 50% per capita GDP percentiles), Linear effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Per Capita GDP Level: Regression 23 <$7,638 Regression 24 <$5,463 β Regression 25 <$3,892 Independent Variable β p value EP/CF Log initial GDP Investment ratio Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 0.156 −6.040 0.161 0.677 −0.495 −0.906 0.000 −0.019 0.490 0.001 0.001 0.229 0.339 0.033 0.949 0.758 0.376 −2.327 0.134 0.351 −0.052 −0.914 0.010 −0.093 0.052 0.008 0.002 0.394 0.894 0.016 0.032 0.059 0.327 −1.908 0.083 0.244 −0.281 −1.341 0.009 −0.042 0.433 0.087 0.004 0.024 0.454 0.363 0.000 0.042 0.325 0.808 0.246 −1.722 0.079 0.403 −0.365 −1.190 0.007 −0.079 −1.241 0.165 0.001 0.021 0.171 0.202 0.000 0.131 0.035 0.227 2.746 0.034 0.681 0.185 0.151 0.884 −1.087 0.216 0.849 0.795 0.656 52.617 0.675 0.463 0.606 0.001 1.206 0.279 0.062 17.722 0.466 0.779 0.963 0.029 −1.531 −1.612 −0.586 19.005 0.164 0.050 0.610 0.001 −1.928 −1.993 −1.505 17.799 0.037 0.009 0.199 0.000 p value β Regression 26 <$2,904 p value β Observations Cases Mean time periods per case 90 24 135 33 180 46 225 58 3.750 4.091 3.913 3.879 R2 within R2 between R2 overall 0.387 0.902 0.570 0.402 0.842 0.578 0.345 0.728 0.463 0.364 0.659 0.484 p value However robust the correlations, inference is a separate endeavor. For example, what single-party dictatorships do, more exactly, with respect to macroeconomic policy that produces higher investment ratios but not growth rates is ambiguous. Also, what democratic governments in more affluent countries are restrained from doing, which increases allocation but not accumulation efficiency, is unclear. In general, these regressions—and those in Kormendi and Meguire (1985), Grier and Tullock (1989), and Barro (1996), for that matter— do not preclude some connection between regime type and monetary or fiscal policy that, in turn, profoundly affects performance.45 More limited time-series analysis or perhaps case studies, as Fischer recommends (1991, pp. 37–38), might better approximate the possible interaction between regime type and particular policy choices. Acknowledgments The author thanks, without implication, Richard R. Nelson, Douglas A. Chalmers, Robert Y. Shapiro, Brendan O’Flaherty, Alessandra Casella, Lauren Paige Kennedy, an 104 DURHAM Table 9. Growth regressions: Developed sample (80%, 70%, 60%, and 50% per capita GDP percentiles), Linear-log effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Per capita GDP Level: Regression 27 >$7,638 Regression 28 >$5,463 Regression 29 >$3,892 Regression 30 >$2,904 Independent Variable β p value β p value β p value β p value Log EP/CF Log initial GDP Investment ratio Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept 0.646 −3.137 0.146 0.017 −0.002 −0.148 0.001 0.006 0.000 0.026 0.000 0.967 0.996 0.735 0.896 0.887 0.216 −1.861 0.115 0.301 0.032 −0.684 0.011 −0.102 0.090 0.028 0.009 0.468 0.934 0.112 0.021 0.047 0.194 −1.515 0.077 0.147 −0.183 −1.235 0.009 −0.041 0.531 0.049 0.009 0.029 0.653 0.554 0.000 0.037 0.326 0.760 0.069 −1.425 0.076 0.386 −0.355 −1.187 0.007 −0.065 −1.258 0.317 0.003 0.022 0.183 0.206 0.000 0.108 0.067 0.202 1.386 0.139 1.017 0.434 −0.368 0.712 −1.277 0.132 −2.065 0.190 −0.155 27.144 0.267 0.809 0.853 0.032 0.467 0.455 −0.425 14.440 0.793 0.645 0.757 0.068 −1.752 −1.586 −0.976 16.322 0.105 0.045 0.376 0.002 −1.949 −1.983 −1.695 15.647 0.027 0.005 0.119 0.000 Observations Cases Mean time periods per case 90 24 135 33 180 46 225 58 3.750 4.091 3.913 3.879 R2 within R2 between R2 overall 0.387 0.963 0.641 0.400 0.831 0.581 0.337 0.761 0.468 0.355 0.670 0.483 anonymous referee, and an associate editor for helpful comments and suggestions regarding this project. Notes 1. More recently, he suggests that “it is naı̈ve to suppose that one of these types of political systems (regimes) will regularly have better economic performance than the other.” New or “transitory” democracies, which might abandon property rights to win elections, are a particular exception to Olson’s benevolent view (Clague et al., 1996, pp. 243–245). 2. This democratic view has strong normative appeal because there is no dilemma between liberty and wealth (Pourgerami, 1991, p. 201). In Bhagwati’s estimation, “all goods things may sometimes go together” (Bhagwati, 1995, p. 50, emphasis added). Of course, one need not actually posit the superior economy efficiency of democracy to be optimistic. To posit no difference between regimes is normatively preferable—democracy should attend economic growth simply because autocracy provides no intrinsic advantage. 3. Jaggers and Gurr review Arat, Bollen, Coppedge and Reinicke, the Gastil Freedom House annual, Gasiorowski, and Vanhanen (Jaggers and Gurr, 1995, pp. 473–476). 4. Brunetti also suggests that this “dominating measure” is “rather subjective” (1997, p. 166). 105 ECONOMIC GROWTH AND POLITICAL REGIMES Table 10. Investment regressions: Developed sample (80%, 70%, 60%, and 50% per capita GDP percentiles), Linear effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (five-year averages), 1960 to 1989. Per capita GDP Level: Regression 31 >$7,638 Regression 32 >$5,463 Regression 33 >$3,892 Regression 34 >$2,904 Independent Variable β p value β p value β p value β p value EP/CF Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept −0.326 −0.720 1.364 −0.189 1.571 0.010 −0.645 0.497 0.801 0.230 0.868 0.051 0.647 0.000 −0.358 −3.395 0.153 0.862 3.529 0.012 −0.252 0.379 0.088 0.859 0.309 0.000 0.458 0.047 −0.968 0.628 0.282 0.778 2.704 0.045 0.070 −9.094 0.026 0.699 0.675 0.257 0.000 0.003 0.521 0.131 −0.471 −0.417 1.269 −0.125 2.374 0.023 −0.115 −6.228 0.235 0.748 0.038 0.842 0.000 0.086 0.198 0.050 −12.839 0.012 −9.853 0.017 −2.994 0.431 −4.558 0.103 −15.001 −3.067 −8.896 34.061 0.040 0.463 0.124 0.229 −13.471 2.017 −4.047 49.140 0.027 0.555 0.420 0.010 −2.586 8.441 1.873 0.948 0.493 0.004 0.703 0.947 −3.665 6.423 0.166 13.731 0.204 0.006 0.970 0.230 Observations Cases Mean time periods per case 91 24 136 33 181 46 226 58 3.792 4.121 3.935 3.897 R2 within R2 between R2 overall 0.669 0.467 0.459 0.572 0.415 0.416 0.413 0.475 0.363 0.303 0.553 0.489 5. For a variety of definitions see Ranney (1968, pp. 143–146). 6. This generally refers to regimes in which a ruling despot, often a monarch, coexists with an elected legislature. 7. LaPalombara and Weiner (1996, p. 6) and Huntington (1968, p. 409) stress this notion. 8. In point of fact, single-party mobilization for development, as under Stalin and Mao, frequently attends extensive party purges. Gerschenkron’s discussion of industrial development under Stalin, one of his exemplars of dictatorial response to economic backwardness, notes Stalin’s suppression of the Communist Party. Jowitt similarly argues that Stalin’s personal rule from 1934 through 1953 threatened the Communist Party as the primary source of charisma (1992, p. 9). 9. Jowitt makes a trajectory argument about the origins of corruption in Marxist-Leninist regimes, by far the most influential exemplar of single-party rule. Briefly, following Weber’s notion of “routinization,” charisma is difficult to sustain as a source of legitimacy, and procedure and routine must ultimately justify rule. The distinction between revolutionary tasks for the party and the security and status of its elite becomes less clear over time, which leads to corruption. 10. Several sources of data are available to compute the number of effective parties. See the Chronicle of Parliamentary Elections and Developments, Inter-Parliamentary Union, Vols. 2–27 (1967–1993): The Political Handbook of the World (various issues); and Machie and Rose (1991). 11. Values are undefined for cases that experience a regime change or a civil war according to the Polity III data set during a five-year averaging period. To some degree, then, regressions that include this measure account for political instability. 106 DURHAM Table 11. Investment regressions: Developed sample (80%, 70%, 60%, and 50% per capita GDP percentiles), Linear-log effective party/constitutional framework hypothesis, GLS random effects models: Medium-run (fiveyear averages), 1960 to 1989. Per capita GDP Level: Regression 35 >$7,638 Regression 36 >$5,463 Regression 37 >$3,892 Regression 38 >$2,904 Independent Variable β p value β p value β p value β p value Log EP/CF Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Dummy for Oceania Intercept −0.199 −0.882 1.420 −0.222 1.467 0.008 −0.665 0.739 0.758 0.223 0.847 0.095 0.693 0.000 0.216 −3.964 0.175 0.855 3.719 0.008 −0.322 0.346 0.036 0.840 0.311 0.000 0.620 0.009 −0.216 −0.359 0.321 0.770 2.712 0.043 0.045 −9.857 0.256 0.820 0.643 0.274 0.000 0.005 0.681 0.106 0.001 −0.893 1.138 0.006 2.409 0.022 −0.126 −6.497 0.994 0.478 0.070 0.993 0.000 0.092 0.156 0.041 −12.588 0.017 −11.287 0.005 −2.702 0.489 −4.877 0.086 −14.622 −3.252 −8.803 35.058 0.058 0.451 0.134 0.217 −15.463 0.455 −5.627 55.788 0.009 0.889 0.248 0.002 −3.013 7.805 1.746 8.412 0.433 0.008 0.727 0.546 −4.213 5.877 −0.123 17.656 0.147 0.012 0.978 0.117 Observations Cases Mean time periods per case 91 24 136 33 181 46 226 58 3.792 4.121 3.935 3.897 R2 within R2 between R2 overall 0.669 0.454 0.447 0.565 0.446 0.447 0.400 0.462 0.357 0.297 0.551 0.488 12. The raw number of effective parties in the legislature correlates quite closely with the effective party/ constitutional framework proxy (.944). 13. Furthermore, twenty-eight of the top thirty cases are democratic according to the Polity III dummy transformation. However incompatible the temporal coverage, twenty-seven of the top thirty countries qualify as democracies using the Gastil measures. 14. For a useful description, see Held (1996). 15. Brief reference to the United States, a “presidential” regime, helps illustrates how the effective party/ constitutional framework measure differentiates democracies. With respect to Congress, the United States is a two-party system, but only one party captures the presidency. Given equal weights regarding the legislative and executive functions of government in the process of economic policymaking, the weighted average number of effective political parties in the United States is approximately 1.5. 16. While never drafted directly into a given constitution, electoral laws have considerable impact on the number of effective parties. For example, in their purest manifestations, the first-past-the-post system tends to exclude smaller parties, and proportional representation leads to a proliferation in parties. The second component of the alternative measure, regarding the relative sovereignty of the executive and legislative government branches, is more directly related to constitutional stipulations. As reforms in Eastern Europe after the cold war illustrate, politicians make distinct choices between different constitutional frameworks. 17. Durham (1998a) discusses the three alternative functional forms. 18. Barro, who uses the Gastil indices, finds support for this functional form (1996, p. 14). 107 ECONOMIC GROWTH AND POLITICAL REGIMES Table 12. Single-party and party-less dictatorships,d GLS random effects models for growth and investment, 1960 to 1989. Growth Regression 39 Investment Ratio Regression 40 Independent Variable β p value β EP/CF Investment ratio Log initial GDP Male education rate Female education rate Population growth Openness to trade Government ratio Dummy for Africa Dummy for Central and North America Dummy for South America Dummy for Europe Intercept 1.054 0.101 −1.297 −0.202 0.039 −0.145 0.015 −0.028 −2.174 0.127 0.045 0.013 0.535 0.916 0.582 0.040 0.466 0.018 2.803 0.022 1.762 0.467 −0.248 0.118 0.075 −0.082 −5.031 0.101 0.431 0.737 0.757 0.000 0.248 0.013 −1.772 0.282 −0.620 0.855 1.147 0.887 9.149 0.438 0.579 0.020 4.678 12.488 −4.395 0.128 0.000 0.590 Observations Cases Mean time periods per case 151 55 2.745 152 55 2.764 R2 within R2 between R2 overall 0.364 0.523 0.463 0.263 0.812 0.753 p value d. Party-less dictatorships that inform the regression estimates include Bahrain (1980–1984), Benin (1970–1974), Bolivia (1970–1974), the Central African Republic (1970–1984), Chile (1975–1989), Ecuador (1965– 1974), Ghana (1985–1989), Haiti (1965–1969), Honduras (1975–1979), Indonesia (1965–1969), Iran (1960–1964, 1985–1989), Iraq (1960– 1964), Jordan (1960–1989), Kuwait (1985–1989), Lesotho (1970–1989), Mali (1970–1974), Myanmar (1965–1969), Nepal (1965–1985), Niger (1975–1984), Pakistan (1960–1964, 1980–1984), Peru (1970–1974), Portugal (1960–1969), Rwanda (1975–1979), Spain (1960–1974), Swaziland (1975–1989), Syria (1965–1969), Thailand (1960–1964), Uganda (1970– 1974), and Uruguay (1975–1979). Single-party authoritarian regimes include Algeria (1965–1989), Benin (1975–1989), Bulgaria (1985– 1989), Cameroon (1965–1989), China (1975–1989), Congo (1980– 1989), Czechoslovakia (1970–1989), East Germany (1975–1984), Hungary (1975–1989), Iraq (1970–1979), Kenya (1965–1989), Liberia (1965– 1984), Malawi (1970–1989), Mali (1980–1989), Mozambique (1980– 1989), Myanmar (1975–1989), Niger (1965–1969), Panama (1970– 1979), Poland (1975–1984), Romania (1965–1984), Rwanda (1985– 1989), Senegal (1965–1974), Sierra Leone (1980–1989), Singapore (1965–1979), Sudan (1975–1984), Taiwan (1960–1969), Togo (1970– 1989), Tunisia (1965–1989), USSR (1965–1989), Yugoslavia (1965– 1989), Zaire (1965–1989), and Zambia (1975–1989). 108 DURHAM 19. This resembles the more explicit “life-cycle” explanations of vertical and horizontal firm integration in neoinstitutionalist economics (see Williamson, 1985, p. 90). Similar to this debate, the former is purportedly more appropriate for early product phases, while the latter is supposedly better suited for advanced stages. 20. Some exceptions might include the popular advocacy of “autonomous” central banks, which purportedly must be insulated from political and popular pressures. But while this view generally advocates discretion, few commentators advocate complete authoritarian rule. 21. With the exception of Helliwell (1992), none of the econometric growth models that include some measure of political regime type use simultaneous equation methods to account for such purported bias. 22. Przeworski and Limongi fail to make this crucial distinction and write, “If democratic regimes are more likely to occur at a higher level of development . . . then regimes are endogenously selected” (1993, p. 62, emphasis added). This is mistaken. Levels of development, an independent (exogenous) variable in neoclassical growth regressions, are quite different from growth rates, the dependent (endogenous) variable, as economists and other social scientists routinely distinguish in growth models. Przeworski et al. (1995) similarly fail to make the distinction. 23. For recent examples, see Londregan and Poole (1996), Przeworski and LImongi (1997, p. 169), and Barro (1996, p. 16). 24. As Przeworski and Limongi succinctly summarize the causal process, “a story told about country after country is that after they develop, social structures become complex, labor processes begin to require the active cooperation of employees, and new groups emerge and organize . . . the system can no longer be run by command” (1997, p. 157). 25. Londregan and Poole’s analysis of “the coup trap” is instructive. Briefly, they find that “the probability that a government is overthrown by a coup d’etat is substantially influenced” by both the rate and level of economic development in both dictatorships and democracies, as underdevelopment and poor performance threaten tenure (Londregan and Poole, 1990, p. 177, emphasis added). Brunetti makes a similar claim about general not particular regime longevity and short-run economic performance (Brunetti, 1997, p. 165), and Alesina et al. (1996) illustrate the vulnerabilities of both democracies and dictatorships to economic stagnation (p. 191). 26. Przeworski et al. (1995) and Przeworski and Limongi (1997) publish no compelling evidence that democracies are more sensitive than dictatorships to poor economic performance. Przeworski suggests that a expected life of a “declining democracy” is eighteen years compared to forty-one years for a “declining dictatorship” (1995, p. 19). But this finding is due to peculiar coding on the dependent variable. Przeworski and Limongi (1997) readily admit this calculation does not include dictatorships that experience transitions to dictatorships. They explicitly write that “we do not distinguish successive dictatorships . . . even if ayatollahs succeed a shah, we treat it as one continuous spell of dictatorship” (1997, p. 160 n.). Therefore, Przeworski (1994, 1995), Przeworski et al. (1995), and Przeworski and Limongi (1997) do not consider coups within a given “continuous spell of military regimes” the end of a regime, as they simply code away the authoritarian vulnerability to economic decline within such “spells.” This evidence does not evaluate the appropriate “survival bias” question of whether the democratic or authoritarian nature of a particular regime makes it more susceptible to a coup, revolution, or some other form of termination under (equally) poor economic performance. 27. Individual short-run time series equations for each country are also instructive, with particular respect to the suspect validity of pooling across a disparate data set. For example, the definition of “effective” party is more consistent within a given country rather than across cases (Durham, 1998a). 28. Time-specific coefficients are available on request. 29. The proxies for Y2 and respective lagged values include the twenty-one-point ordinal Polity III scale, a related dummy variable for “coherent” democracy and the effective party/constitutional framework measure. Turning to the Z set, the international considerations of the number of democracies inside and outside a given continent from 1960 to 1989 are defined according to Stepan and Skach (1993) and the Polity III data set. Initial levels of development and growth rates are from the Penn World Table 5.6. Barro and Lee’s updated data set on average years of schooling for the population over twenty-five proxies male and female education levels from Barro and Lee (1993). Past and contemporary membership in the Commonwealth defines the British colonial dummy from the Political Handboook of the World (1992, pp. 909–910). 30. The regressions on regime type proxies are available on request. 31. The sensitivity analysis in Durham (1998a) also includes alternative specifications that include or exclude the log and level of development and the education rate. 32. These models, which follow the sample division in Przeworski et al. (1995), are also available on request. ECONOMIC GROWTH AND POLITICAL REGIMES 109 33. The simple correlation statistic between the effective party/constitutional framework measure and the initial log GDP per capita is 0.572. 34. The following uses this term somewhat loosely. “Neoclassical” equations include variables, such as investment in human capital and openness to international trade, more commonly associated with “new” or “endogenous” growth theory. 35. Brunetti similarly notes that investigations of political regime type are particularly underspecified with respect to common macroeconomic determinants (1997, p. 165). 36. Barro and Sala-i-Martin (1995), Barro (1996), and Perotti (1996) report separate coefficients for male and female education rates. The theoretical justification for differentiating by sex is that “the gap . . . is an indicator of an economy’s backwardness and that greater backwardness induces a higher growth rate through the familiar convergence mechanism” (Barro, 1996, p. 6). 37. The source for these data is Summers and Heston’s Penn World Table 5.6, except education rates, which follow Barro and Lee (1993). See Fischer for some discussion of the validity of the data set (1993, p. 495). For a similar discussion of government spending and trade see Durham (1998a, 1998b). 38. Following national income accounts, government consumption of GDP is the measure. This proxy unfortunately does not distinguish infrastructure investment from regulation. 39. Exports and imports as a fraction of GDP from the Summers and Heston data set measures the concept. 40. For example, Katzenstein (1985) suggests that economies of scale and their position in the international economy forced smaller advanced European states to pursue export-oriented strategies, as international influences stimulated consensus and restraint between labor and capital domestically. 41. Collinearity between the effective party/constitutional framework measure and the initial level of development affects few results. Regressions for each specification that do not control for “convergence” produce the same results save two equations. The log number of effective parties is positively significant at the 10 percent level in total sample investment equation (Regression 5), and the quadratic form of the number of effective parties is significant at the 10 percent level in the total sample investment equation (Regression 6). These results are available on request. 42. As the preceding section on the supposed endogenous selection between political regime type and growth suggests, more advanced countries are generally more “democratic.” For example, among the sample of countries above the mean per capita GDP of $4,315, the effective party/constitutional framework mean is 2.4, and the raw effective party mean is 2.5. Also, 74 percent of this sample qualify as “coherent” democracies according to the Polity III scale. Conversely, countries below the mean are more “authoritarian,” with an effective party/constitutional framework mean of 1.1 and raw effective political party mean of 1.3. Only 17 percent of this sample qualify as democratic. 43. Given the unidirectional nature of the hypothesis, the following does not consider the quadratic form. 44. Party-less regimes represent a disparate performance category, with a higher pooled standard deviation of growth, 3.9 percent, compared to single-party regimes, 3.5 percent, from 1960 to 1989. But both subtypes of dictatorships exhibit greater volatility than multiparty systems, which show a pooled standard deviation of growth of approximately 2.7 percent during the period. 45. Similarly, Helliwell concludes that there is an understudied link between regime type and policymaking (1992, p. 23). References Alesina, A., et al. (1996). “Political Instability and Economic Growth.” Journal of Economic Growth 1, 189–211. Bardhan, P. (1994). “Floor Discussion of ‘The Impact of Constitutions on Economic Performance,” by Elster.” Proceedings of the World Bank Annual Conference on Development Economics (pp. 236–239). Washington: IBRD/World Bank. Barro, R. (1996). “Democracy and Growth.” Journal of Economic Growth 1, 1–27. Barro, R., and J. W. Lee. (1993). “International Comparisons of Educational Attainment.” Journal of Monetary Economics 32, 363–394. Barro, R., and X. Sala-i-Martin. (1995). Economic Growth. New York: McGraw-Hill. Bhagwati, J. (1995). “The New Thinking on Development.” Journal of Democracy 6, 50–64. 110 DURHAM Brunetti, A. (1997). “Political Variables in Cross-Country Growth Analysis.” Journal of Economic Surveys 11, 163–190. Chronicle of Parliamentary Elections and Developments. (1967–1993). Inter-Parliamentary Union. Vols. 2–27. Clague, C., et al. (1996). “Property and Contract Rights in Autocracies and Democracies.” Journal of Economic Growth 1, 243–245. DiPalma, G. (1990). To Craft Democracies: An Essay on Democratic Transition. Berkeley: University of California Press. Durham, J. B. (1998a). “Economic and Political Institutions and Growth: Some Econometrics.” Ph.D. diss., Columbia University. Durham, J. B. (1998b). “Economic Growth and Institutions.” Under review. Duverger, M. (1954). Political Parties. New York: Wiley. Fischer, S. (1993). “The Role of Macroeconomic Factors in Growth.” Journal of Monetary Economics 32, 485–512. Fischer, S. (1991). “Growth, Macroeconomics, and Development.” NBER Working Papers No. 3702. Fischer, S. (1987). Freedom in the World: Political Rights and Civil Liberties, 1986–1987. New York: Freedom House. Gerschenkron, A. (1962). Economic Backwardness in Historical Perspective: A Book of Essays. Boston: Harvard University Press. Gourevitch, P. (1978). “The Second-Image Reversed: The International Sources of Domestic Politics.” International Organization 32, 881–912. Grier, K., and G. Tullock. (1989). “An Empirical Analysis of Cross-National Economic Growth, 1951–1980.” Journal of Monetary Economics 24, 259–276. Held, D. (1996). Models of Democracy. Palo Alto: Stanford University Press. Helliwell, J. F. (1992). “Empirical Linkages Between Democracy and Economic Growth.” NBER Working Papers No. 4066. Huntington, S. (1968). Political Order in Changing Societies. New Haven: Yale University Press. Jaggers, K., and T. R. Gurr. (1995). “Tracking Democracy’s Third Wave with the Polity III Data.” Journal of Peace Research 32, 469–482. Jowitt, K. (1992). New World Disorder: The Leninist Extinction. Berkeley: University of California Press. Katzenstein, P. (1985). Small States in World Markets: Industrial Policy in Europe. Ithaca: Cornell University Press. Kohli, A. (1986). “Democracy and Development.” In J. P. Lewis and V. Kallab (eds.), Development Strategies Reconsidered. New Brunswick: Transaction Books. Kormendi, R., and P. Meguire. (1985). “Macroeconomic Determinants of Growth.” Journal of Monetary Economics 16, 141–163. Kurth, J. (1979). “Industrial Change and Political Change: A European Perspective.” In D. Collier (ed.), The New Authoritarianism in Latin America. Princeton: Princeton University Press. Laakso, M., and R. Taageper. (1979). “Effective” Number of Parties: A Measure with Application to West Europe.” Comparative Political Studies 12, 3–27. LaPalombara, J., and M. Weiner. (1966). Political Parties and Political Development. Princton: Princeton University Press. Levine, R., and D. Renelt. (1992). “ A Sensitivity Analysis of Cross-Country Growth Regressions.” American Economic Review 82, 942–963. Lipset, S. M. (1959). “Some Social Requisites of Democracy: Economic Development and Political Legitimacy.” American Political Science Review 53, 69–105. Londregan, J., and K. Poole. (1990). “Poverty, the Coup Trap, and the Seizure of Executive Power.” World Politics 42, 151–183. Londregan, J., and K. Poole. (1996). “Does High Income Promote Democracy?” World Politics 49, 1–30. Machie, T., and R. Rose. (1991). The International Almanac of Electoral History: Fully Revised Third Edition. Washington Congressional Quarterly. Marsh, R. (1988). “Sociological Explanations of Economic Growth.” Studies in Comparative International Research, 41–76. Moore, B. (1966). Social Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern World. Boston: Beacon Press. North, D., and R. Thomas. (1973). The Rise of the Western World: A New Economic History. Cambridge: Cambridge University Press. ECONOMIC GROWTH AND POLITICAL REGIMES 111 North, D., and B. Weingast. (1989). “Constitutions and Commitment: The Evolution of Institutions Governing Public Choice in Seventeenth-Century England.” Journal of Economic History 49, 803–832. Olson, M. (1993). “Dictatorship, Democracy, and Development.” American Political Science Review 87, 567–576. Perotti, R. (1996). “Growth Income Distribution, and Democracy: What the Data Say.” Journal of Economic Growth 1, 149–187. Political Handbook of the World 1992. (1992). New York: McGraw-Hill. Pourgerami, A. (1991). “The Political Economy of Development: An Empirical Examination of the Wealth Theory of Democracy.” Journal of Theoretical Politics 3, 189–211. Przeworski, A. (1994). “Comment on ‘The Impact of Constitutions on Economic Performance.”’ Proceedings of the World Bank Annual Conference on Development Economics (pp. 227–231). Washington: IBRD/World Bank. Przeworski, A. (1995). “The Role of Theory in Comparative Politics: A Symposium.” World Politics 48(1) (October), 1–49. Przeworski, A., and F. Limongi. (1993). “Political Regimes and Economic Growth.” Journal of Economic Perspectives 7, 51–69. Przeworski, A., and F. Limongi. (1997). “Modernization: Theories and Facts.” World Politics 49, 155–183. Przeworski, A., et al. (1995). “What Makes Democracies Endure?” Journal of Democracy 6, 39–55. Ranney, A. (1968). “The Concept of ‘Party.”’ In O. Garceau (ed.), Political Research and Political Theory (pp. 143–146). Boston: Harvard University Press. Rao, V. (1984–1985). “Democracy and Economic Development.” Studies in Comparative International Development 19, 67–81. Romer, P. (1993). “Idea Gaps and Object Gaps in Economic Development.” Journal of Monetary Economics 32, 543–573. Schwarz, G. (1992). “Democracy and Market-Oriented Reform: A Love-Hate Relationship?” Economic Education Bulletin 32. Scully, G. (1988). “The Institutional Framework and Economic Development.” Journal of Political Economy 96, 652-662. Sirowy, L., and A. Inkeles. (1990). “The Effects of Democracy on Economic Growth and Inequality: A Review.” Studies in Comparative International Development 25, 126–157. Stepan, A., and C. Skach. (1993). “Constitutional Frameworks and Democratic Consolidation: Parliamentarism Versus Presidentialism.” World Politics 45, 1–22. Summers, R., and A. Heston. (1991). “The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950–88.” Quarterly Journal of Economics 61, 327–368. Von Der Mehden, F. (1964). Politics of the Developing Nations. Englewood Cliffs: Prentice-Hall. Weede, E. (1983). “The Impact of Democracy on Economic Growth: Some Evidence from Cross-National Analysis.” Kyklos 86, 21–39. Williamson, O. (1985). The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting. New York: Free Press. Wittman, D. (1989). “Why Democracies Produce Efficient Results.” Journal of Political Economy 97, 1395–1424. View publication stats