In “Fitzgibbon Survey of Latin American Democracy: An Update of the 2000 Tabulations,” Emporia State University Political Science Professor Phil Kelly asked whether “certain reformist and/or radical states, such as Cuba and Nicaragua, [are] given higher scale rankings because the majority of survey participants reflect a ‘liberal’ bias as was seen in the 1985 Johnson-Kelly Attitudinal Profile?” (Kelly 2003, 2).2 However, Kelly, three-time administrator of the quinquennial canvassing of “scholarly images of democracy” in the region (1991, 1995, and 2000), did not answer his own question. Instead, he proceeded to show that the Survey’s democracy rankings correlate more or less strongly with other democracy indicators and that nearly 90% of the variance in “the 1945–2000 cumulative democracy rankings” is accounted for by two measures of development: daily newspaper circulation per capita and tractors per hectare (Kelly 2003, 4). Again, left out of the analysis was any statistical test that would suggest an answer to the question of whether there is a leftist bias in the “images” of Latin American democracy generated by the Survey. The purpose of this research note is to see whether such a pattern can be detected in the data.
THE FITZGIBBON-JOHNSON LATIN AMERICAN DEMOCRACY INDEX
The Fitzgibbon-Johnson Index, named after its originator, UCLA political scientists Russell H. Fitzgibbon and Kenneth Johnson, Fitzgibbon’s student and immediate successor as administrator of the Survey after the latter’s retirement, is constructed from responses to canvasses of Latin Americanists conducted every five years. The first questionnaire was sent to “a panel of ten distinguished latinamericanist (sic) scholars” in 1945 (Kelly 2003, 1). They were asked to assess the “strength” of democracy in Latin American countries using 15 criteria, to wit: educational level, standard of living, social legislation, internal unity, political maturity, civilian supremacy, judicial independence, probity in the management of government funds, professionalism in public administration, autonomy of local government, and five freedoms (of elections, party organization, and the press, and from foreign and ecclesiastical domination).3 Altogether, “without significant adjustment” to the instrument (Kelly 2003, 2), thirteen surveys have been conducted, the last one administered by Joseph Klesner of Kenyon College in 2005. The number of participating Latin Americanists, however, has multiplied, with nearly 80 taking part in the 2005 Survey. (I myself participated in 2005 and at least one prior survey.)
As the 15 criteria noted above indicate, Fitzgibbon- Johnson embodies what might be called a “maximalist” conception of democracy. That is, it goes well beyond strictly institutional features of regimes to include social (educational level), economic (standard of living), and even ideological or at least rather subjective considerations (social legislation, freedom from foreign or ecclesiastical domination).
By contrast, in Democracy and Development, Przeworski et al. (2000) offer a “minimalist” conception. A regime is considered a democracy if the legislature and the executive are elected (the latter either directly by the electorate or by the legislature), there is more than one party with an ex-ante chance of winning elections, and there is alternation in office. If the last feature is not observed, the regime is classified as a dictatorship. As Przeworski (n.d., 3) puts it, “Operationally, a democracy is a regime in which incumbents lose elections and leave office if they do.”
In-between these two extremes lies the POLITY IV regime classification. In this typology a regime is assigned a yearly value of between 0 and 10 on two scales simultaneously, a democracy scale and an autocracy scale. Then the value of the latter is subtracted from the former to obtain a polity score for any given year. A regime is more or less autocratic or democratic depending on how the chief executive is chosen, how constrained and accountable he is, and how open and competitive is the degree of political participation (Marshall and Jaggers 2002). A perfect democracy receives a score of 10 and an absolute autocracy, -10.
Although their classifying criteria are somewhat different, what Prezeworski et al. and POLITY IV have in common is that, unlike Fitzgibbon-Johnson, they rate regimes exclusively on the basis of institutional features, omitting other considerations altogether. Thus, they provide a useful reference point in the search for ideological biases in the latter’s rankings.
CASTRO’S CUBA DEMOCRACY RANKINGS, 1960–2000
Of particular relevance to Kelly’s question about a leftist bias among Latin Americanists is the “image” of Castro’s regime yielded by the Survey. Table 1 compares that regime’s “democracy” rankings in Fitzgibbon-Johnson not only with those of Przeworski et al. and POLITY IV, but with several others included in Kelly’s 2003 “Update.”4 Where I have found more recent values in any of the indexes, as in those for Polyarchy and POLITY IV, these have been entered. Note that except in 1960 and 1965, Cuba ranks higher on democracy in the Fitzgibbon- Johnson than with any other measure. Across all years, the mean democracy rank for the Castro regime is 13 in Fitzgibbon-Johnson and 17.9 in all other indexes combined. The difference is even greater in the years 1985–2000: 13.3 vs.19.5. In short, whereas other indexes rank Castro’s regime at or near the bottom, the Latin Americanists place it onefourth to one-third higher on the scale. Although additional factors could certainly be adduced, it would be unreasonable to dismiss out of hand the possibility that the discrepancy is suggestive of an ideological bias, especially given the well-documented leftism of the resolutions of the Latin American Studies Association (Cuzán 1994, 1995).
A STATISTICAL TEST OF LEFTIST BIAS IN THE FITZGIBBON SURVEY RESULTS
Cuba is not the only country which ranks substantially different in Fitzgibbon-Johnson than in other assessments of democracy. Taking as our benchmark the cumulative rankings obtained with POLITY IV across the same years for which there are Fitzgibbon Survey data (starting with 1960, when the Castro regime made its first appearance, through 2000), for every country I subtracted the cumulative value of POLITY IV from that of Fitzgibbon-Johnson. If the residual is negative it means that the latter over-rates a country’s democracy relative to the former and vice-versa for a positive number.
As well as Cuba, five other countries are rated substantially more or less democratic in Fitzgibbon- Johnson than is called for by their institutional features as measured by POLITY IV. (By “substantial,” I mean that the difference in the rankings is at least 1.5 standard deviations away from the mean absolute total difference between the two indexes.) Shown in Figure 1, these are Argentina, Bolivia, Guatemala, Honduras, and Mexico.
What might account for these discrepancies? Two factors come to mind. One is development. Argentina is one of the most economically developed countries in Latin America, while Bolivia and Honduras are among the least developed. Since, as we have seen, Fitzgibbon-Johnson explicitly takes into account economic development as one of the criteria for assessing the strength of democracy, it is not surprising that Argentina’s democratic quality is overrated and that of Bolivia or Honduras are under-rated relative to POLITY IV.
As suggested in the previous section, ideological bias may also be a factor. Throughout the entire 1960– 2000 period, Cuba was ruled by, to borrow Kelly’s terminology, the most “radical” regime in the region. Until the 1988 election, Mexico was ruled by (again borrowing from Kelly) a “reformist” or leftist dominant- party regime of the corporatist type, one that was officially anti-clerical, rhetorically populist and revolutionary, and heavily involved in managing large swaths of the economy. These two regimes, then, would be expected to enjoy the ideological sympathy of “liberal” academics. By contrast, Guatemala was for most of the period ruled by regimes— some elected, some military—engaged in a decadeslong counter-insurgency campaign against “leftist” guerrillas. Hence, Guatemalan governments, even those composed of civilians winning office in competitive elections, would be expected to be antipathetic to “liberal” academics. These expectations are borne out. Relative to POLITY IV, Cuba and Mexico are over-rated in Fitzgibbon-Johnson and Guatemala is under-rated as democracies.
Although suggestive, these few cases are insufficient to support the hypothesis of a leftist bias in the Survey. A more reliable statistical test is needed. To that end, as shown in Table 2, for the years 1960– 2000 several dependent variables were constructed from Fitzgibbon-Johnson as well as from POLITY IV and Przeworski et al.5 The dependent variables, FSUMR-P4RAWR, FSUM-POL4SUM, FSUMR-POL4SUMR and FSUMR-PRZER, all measure differences between the country democracy rankings obtained with Fitzgibbon-Johnson and either of the other two indexes.
Next, I regressed each of these variables on two predictors. One is the 1970–2000 average GDP per capita in 1990 dollars.6 This is a measure of development which, as we have seen, is explicitly taken into account in Fitzgibbon-Johnson’s “maximalist” conception of democracy. Although other measures of development or combinations thereof could have been used, per capita income is probably the best single measure.
The other predictor variable is LEFTISM, scored as follows. Regimes that are “radical” or “reformist” are assigned a score that is a multiple of -1 times the fraction of Survey years between 1960 and 2000 that the country was under its rule. Thus, Cuba was assigned a value of -1, Mexico -0.66 and Nicaragua -0.22. The rationale for Mexico’s value is that the regime was leftist only between 1960 and 1985. After the 1988 election, the ruling Partido Revolucionario Institucional began to implement a policy about-face, establishing diplomatic relations with the Vatican, privatizing state enterprises, negotiating a free trade agreement with the United States and Canada, etc. (Raising Mexico’s leftism score would only strengthen the effect of this variable.) As for Nicaragua, it was ruled by the Sandinistas’ leftist regime for two of the nine Survey years, 1980 and 1985, having been ousted in the 1990 election.
On the other hand, regimes engaged in a counter-insurgency campaign against leftist guerrilla movements eliciting widespread sympathy from abroad are assigned a score of 1 times the fraction of Survey years between 1960 and 2000 that the campaign lasted. Accordingly, Guatemala, El Salvador, and Peru are respectively scored 0.78 (7 out of 9 Survey years), 0.44 (4 out of 9), and 0.22 (2 out of 9). All other countries are scored 0 on this variable.
Table 3 displays the results of the statistical analysis. To interpret these results correctly, it is important to remember what the dependent variables measure. Each one is constructed by subtracting the POLITY4 or Przeworski et al. democracy values from those of Fitzgibbon-Johnson. Since the countries are ranked from 1 to 20, with 1 being the most democratic, a negative residual means that Survey respondents over-rated the country’s democracy relative to one or the other measure and vice-versa for a positive residual.
Note that both independent variables are statistically significant and behave as expected. The relationship between per capita GDP and each of the dependent variables is negative. That is, the higher the country’s income, the more Fitzgibbon-Johnson over-rates the country’s democracy relative to the other measures, while the opposite is true for low-income countries (recall the contrast between Argentina and Bolivia in Figure 1).
Also, the relationship between LEFTISM and each of the dependent variables is positive. That is, countries in which the regime engaged in a counter-insurgency campaign against leftist guerrillas, scoring greater than 0 on this variable, were under-rated on democracy by Fitzgibbon-Johnson relative to the other two measures; conversely, countries ruled by a leftist regime, scoring less than 0 on this variable, were overrated. On average the models account for about 60% of the variance in the difference in the rankings between Fitzgibbon-Johnson, on the one hand, and either POLITY4 or Przeworski et al. LEFTISM alone explains about half of the total.
Incidentally, removing Cuba from the analysis hardly alters the size of the coefficients, although in the model for Variable X, the significance level of LEFTISM goes down from 0.05 to 0.10. The effect of this variable, then, is not solely a function of Cuba’s inclusion. The bias is more generalized than that. Neither does adding a measure of the Gini Index of Inequality make much difference. I borrowed the data from Milanovic and Yitzhaki (2002). Their series has missing values for Cuba, Guatemala, and Haiti. I estimated the values for the first two but not knowing what to do about Haiti, I omitted it from the analysis that follows. For Guatemala I entered 0.50 (the same as that of Bolivia, El Salvador, and Nicaragua, which is a little higher than that of Peru). For Cuba, I entered 0.32, the average of the formerly communist ruled states of Eastern Europe and the former Soviet Republics. This is almost three standard deviations below the Latin American mean (with Cuba, mean=0.49, s.d.=0.07; without Cuba, mean=0.49, s.d.=0.05). The results: the coefficient has the wrong sign and is not statistically significant. Thus, the idea that it is not LEFTISM per se but some notion of equity in income distribution that is behind the observed bias, as was suggested to me by a fellow Survey participant, is not supported.
The evidence analyzed in this paper is consistent with the hypothesis that a leftist bias influences the Fitzgibbon-Johnson democracy ratings. In the case of Cuba, the bias leads to what are probably the most anomalous results of all. In the 2005 Survey (not included in the regressions reported in the previous section), the longest-lasting and arguably one of the most repressive, politically divisive, destructive (economically, socially, culturally), and war-mongering7 dictatorships in the history of Latin America is rated as more “democratic” than seven other regimes in the region (Honduras, Ecuador, Nicaragua, Paraguay, Bolivia, Nicaragua, and Haiti). Leaving aside Haiti, which one could argue should not even be included in the Latin American group at all, over the last decade and a half or so all the other countries ranking lower than Cuba in the Survey have held competitive elections, and of those all but Paraguay have experienced at least one transfer of power from the loser to the winner of an election. In other words, all but one have passed Przeworski et al.’s alternation test for qualifying as a democracy. Yet Cuba, which for nearly half a century has been under the rule of a tyrannical regime, is rated as more “democratic” than they. How could this be?
A close look at the 2005 results yields information on the potential sources of this bias. As noted earlier, Fitzgibbon-Johnson includes 15 separate criteria for assessing a regime’s democratic qualities. Some are purely institutional features of democracy, namely freedom of elections, freedom of the press, freedom to organize political parties, judicial independence, civilian control of the military, and local autonomy. On all but one of these variables, the 2005 Fitzgibbon Survey participants were clear-eyed, ranking Cuba last or next to last (alternating with Haiti). The one exception was civilian supremacy, where the country was ranked 17th, ahead of Paraguay, Guatemala, and Haiti. This is hard to explain, given the well-known militarization of society and the fact that the Castro brothers are hardly ever seen in anything but a military uniform. In any case, 17th is near the bottom, so it’s a small exception.
But Cuba ranked above the median on variables subject to ideological influences or preferences,8 namely freedom from foreign domination (where Cuba is ranked 5th), freedom from ecclesiastical domination (6th), and social legislation (2nd). Additionally, the country was rated 4th on “internal unity,” behind only Costa Rica, Uruguay, and Chile, ahead of Mexico (7th) and Argentina (8th). Finally, the country placed 4th on educational level. This is, of course, one of those areas in which the Castro regime propagandizes ad nauseam, claiming to have made great strides. But given the scarcity of books, the politicization of schools and universities, and the inability of all but trustees of the regime to have access to the internet, 9 such claims cannot be accepted at face value. 10
In sum, in 2005 at least, the Fitzgibbon Survey participants were not in any doubt concerning the dictatorial nature of the Castro regime. Nevertheless, on the strength of the values assigned to variables that are subject to ideological preferences, Cuba ended up ranking 13th overall, up two places from the previous Survey. Although research on the earlier years needs to be done to see if the same pattern holds, it may very well be that the faulty Cuba rankings may have less to do with the ideological orientation of the participants than with the Survey instrument itself. By adopting a maximalist conception of “democracy,” with institutional, political, social, economic, and ideological variables all thrown in and weighted equally, Fitzgibbon-Johnson is liable to produce some rather cloudy “images” of Latin American democracy.
Fortunately, this defect could be easily remedied. The strictly political or institutional variables, five or six in all, could be weighted more heavily. Alternatively, two indexes could be calculated from the same Survey, the usual one and a “minimalist” one that focuses only on the institutional features of regimes. It remains for future research to reveal whether a recalibration of the Index along these lines purges the results obtained with the historical data of what appears to be a leftist bias in the rankings.
1. I sent earlier drafts of this paper to all 2005 Fitzgibbon Democracy Survey participants whose e-mail addresses could be found, asking for feedback. Here I thank those who responded with constructive criticisms, suggestions, or encouragement. Also, I thank Phil Kelly and Joseph Klesner, who graciously commented on earlier drafts of this paper, and Juan del Aguila, who served as discussant at a panel in the 2006 meeting of the Association for the Study of the Cuban Economy where I presented this paper. Thanks, finally, to Dani Manjikian, Office Manager in the Department of Government at UWF, who proofread and assisted in the management of most of the data analyzed herein.
2. In e-mail communication (7/20/2006), Prof. Kelly reports that about 80% of the 1985 Fitzgibbon Survey respondents described themselves, some in highly idiosyncratic terms, as on the liberal or leftist side of the ideological spectrum, and only 12% on the rightist or conservative side.
3. For a description of these criteria, see “Fitzgibbon Survey of Scholarly Images of Democracy in Latin America.” Available: www2.kenyon.edu/Depts/PSci/Fac/klesner/fitzgibbon/default.htm.
4. Kelly included Gastil’s measures for 1978 and 1988. For those, I have substituted the Freedom House Political Rights Index values, though noting the Gastil values Kelly used in his paper in footnotes to Table 1.
5. Przeworski et al. classify every regime year as either a dictatorship (0) or a democracy (1). Their data series ends in 1990. Following their method, I entered values for 1995 and 2000. However, I made a slight adjustment to their operationalization, scoring democratic years 1, dictatorship years -1, and years when a transition took place, 0.
6. Source: United Nations Statistics Division, http://unstats.un.org/unsd/snaama/dnllist.asp.
7. A fellow Survey participant wondered at my use of the phrase “war-mongering.” I pointed out that during the first three decades of life the Castro regime hosted and trained guerrillas from three continents. Also, it sent military advisors to some of these groups and to the Sandinista regime in Nicaragua, as well as thousands of troops to Africa in support of leftist regimes in Angola and Ethiopia, where they incurred substantial casualties. The regime called it “making revolution.”
8. I wrote this paper without consulting my copy of the Survey I filled out. For all I know, I may have been guilty of the very same thing I am criticizing. Which, if this were the case, would support the notion that the fault lies not so much in the ideological orientation of Survey participants as with the instrument itself, which requires respondents to rate countries on irrelevant variables.
9. At 1.1 internet users per 100 population, Cuba scores second lowest in Latin American on this indicator, barely above Haiti (0.96/ 100 population). See Boston University, “The Project for Human Development.” Available: http://humandevelopment.bu.edu/ dev_indicators/start.cfm?header_id=13. Downloaded 7/09/06.
10. Besides, other countries such as Paraguay have made even more impressive literacy gains without receiving much recognition, let alone ideological credit.