CODING HISTORICAL QUALITATIVE CASES: 1979 AS A BENCHMARK YEAR

An extended part of Roberts’s response to my review focuses on my criticisms of his classification of cases into the labour-mobilizing and elitist categories. We overwhelmingly agree about the specifics of cases. We have two major disagreements about how to code cases based on the underlying data.

The first is about whether 1979 needs to be a uniform benchmark for coding cases for the region. I completely agree that in principle, 1979 does not need to be a uniform benchmark for the entire region. However, for reasons related to the measurement of economic crisis in Changing Course, either 1979 needs to be a uniform benchmark, or other parts of the data and argument need to be changed.

The primary evidence for the claim that labour-mobilizing cases were associated with more severe economic crises is in Table 5.1 of Changing Course (Roberts Reference Roberts2014: 94). This table presents data on countries’ peak inflation rate from 1970 to 2000, the number of years of inflation greater than 100 per cent from 1970 to 2000, the worst economic contraction from 1980 to 2000, and change in the real minimum wage from 1980 to 1997. All four summary indicators are measured as effects for the whole period from 1970 or 1980 until 1997 or 2000. The classification of party systems in Table 5.1 and the discussion treat each country case as static for the 1980–2000 period (each country is assigned to the labour-mobilizing or elitist category for the entire period).

In his response to my review, Roberts correctly writes that the ‘1979 benchmark fails to take into account ongoing evolutionary patterns in party systems during the 1980s’. However, if we code cases based on ‘evolutionary patterns . . . during the 1980s’ and measure economic crisis for the entire 1980–2000 period, we implicitly posit that party system dynamics later in the decade help explain variance in the depth of economic crisis starting earlier in this period. If we code party systems based on characteristics after 1980, the classification (e.g. labour-mobilizing party system) comes in part temporally after the economic effects that the labour-mobilizing system are supposedly associated with. In his response, Roberts uses the strong showing of the Partido dos Trabalhadores (PT – Workers’ Party) in Brazil in 1989 to help justify coding it as a labour-mobilizing case. But for a labour-mobilizing party system to affect economic results beginning in 1980, the features of the system must be in place by 1979.

One reason that I focused on 1979, then, is that one of the key outcomes (economic performance) that Roberts associated with the different kinds of party systems is measured from 1980 to 2000. My criticisms of Roberts’s classifications were therefore based on party system characteristics around 1979 or, in cases of dictatorships that banned parties (Argentina, Bolivia, Chile, Peru and Uruguay), at the time of the antecedent military coups (1976, 1964, 1973, 1968 and 1973, respectively).

If Roberts wants to argue for country-specific years for coding party systems, as he does in his response, this should be a consistent choice through Changing Course. It is not. For example, although he coded the peak years of neoliberal reform in country-specific ways, he measured the effects of neoliberal reform (the aftermath) for all countries in the 2000–10 period. Likewise, he coded the end of the transition period (from import substitution industrialization to neoliberalism) as 2000 for all countries, although he acknowledges substantial differences in the timing of neoliberal reform (Roberts Reference Roberts2014: 113–14, Table 6.1). As a result, for the transition period (1980–2000), the book measures economic crisis after the period of peak neoliberal reform had ended in Chile (1979), whereas for Brazil these measures start nearly two decades before the end of the peak period of neoliberal reform (1999).

CLASSIFYING CASES: CHANGE AND CONTINUITY

Roberts and I agree the party systems of Bolivia, Brazil, Mexico, Nicaragua, Peru, Uruguay and Venezuela at the end of the 1970s were very different from the systems at earlier times during the import substitution industrialization era. We disagree about how to handle classification and how to conceptualize change and continuity given these changes over time.

Although we agree on the descriptions of cases, I am sceptical that they all (except perhaps Nicaragua in his interpretation) exhibited sufficient continuity over time to treat them as having essentially the same characteristics on his dimensions (a major left-of-centre party with strong connections to the labour movement and high union density) throughout the import substitution industrialization era. In terms of their origins, the party systems of Mexico and Venezuela were labour-mobilizing in practice, and they meet the criteria Roberts uses to classify such cases. We agree that by the 1970s, in both countries, labour was mostly demobilized and subordinate to parties and/or the state (e.g. Roberts Reference Roberts2014: 78, 95, 215, 218).Footnote ¹ Calling a party system ‘labour-mobilizing’ when it no longer exhibited the dynamics that would lead one theoretically to expect deeper economic crises (e.g. greater labour resistance, a major party that supports labour) is questionable. If we classify the party systems of Mexico and Venezuela as of 1979, they were labour-demobilizing.

We also agree that the dynamics of Uruguay’s party system changed profoundly over time; the emergence of the leftist Frente Amplio (FA – Broad Front) as a major electoral contender in 1971 was an important inflection point in this respect. In 1971, Uruguay had a more powerful, independent and leftist labour movement than Mexico or Venezuela, and unlike the latter two countries, it also had an electorally important leftist party. If we treated the distinction between labour-mobilizing and other systems as a continuous variable, by 1971, I would score the Uruguayan case as considerably closer to an ideal type labour-mobilizing system than the Mexican or Venezuelan.

Coding all cases except Nicaragua as a continuity from roughly 1940 until roughly 1980 is particularly problematic for four party systems that in 1979 diverged sharply from the characteristics that Roberts attributed to labour-mobilizing or elitist party systems: Argentina, Brazil, Chile and Uruguay. Changing Course does not address how profound disruptions of party systems such as those occasioned by dictatorships affect his coding of cases. It treats these cases as continuities from the earlier democratic periods even if the conditions that determine the classification of party systems as labour-mobilizing changed radically. In 1979, Argentina, Brazil, Chile and Uruguay had military dictatorships. The dictatorships in Argentina, Chile and Uruguay had suppressed parties, and in Brazil, until 1980, it allowed only an official governing party and a moderate opposition party. All four dictatorships smashed labour movements, although by the late 1970s the Brazilian labour movement was reviving in Greater São Paulo.

Treating these cases as labour-mobilizing even under military dictatorships has important implications for some claims in Changing Course. As noted above, Roberts (Reference Roberts2014: Table 5.1) measures the economic effects of party systems beginning in 1980. But in Argentina, Brazil and Chile, features of labour-mobilizing systems cannot explain bad economic performance until roughly the founding democratic elections (1983, 1985, 1989, respectively).

Part of my point in signalling deep change in these cases before the 1980s was that many party systems in the region were profoundly reshaped even in the absence of a deep change in economic models. If many party systems exhibited profound change before the beginning of the ‘critical juncture’, we need to question the status of the turn to neoliberal policies as the prime mover of change in party systems.

In his response, Roberts objects to my statement that his categorization of cases (labour-mobilizing versus elitist party systems) was static. As my original review indicated, I viewed his classification of cases, but not his analysis, as static. By static, I meant that once a party system was established as labour-mobilizing or elitist, it remained in that broad category until the disruptions caused by the transition from import substitution industrialization to neoliberalism. Changing Course coded 15 of 16 country cases (with the partial exception of Nicaragua) as belonging to one category or the other and as remaining in that category throughout the period from the 1940s until 1979 (Roberts Reference Roberts2014: 72–3).

EVIDENCE TO SUPPORT THE ARGUMENT

My review praised the ambition and depth of knowledge manifest in Changing Course, but it also concluded that the evidence did not convincingly support the argument that labour-mobilizing systems generated deeper economic crises and more party system instability than elitist systems during the transition from import substitution industrialization to neoliberalism. For the argument to be convincing, we should expect three conditions to be met for both the profundity of economic crisis and of party system change (the main outcome variable) in Chapters 5, 7 and 8.

Firstly, during the transition from import substitution industrialization to neoliberal economic reforms, indicators of economic crisis and of party system instability should on average be higher (by a statistically significant percentage) in the labour-mobilizing systems. Based on the evidence in Changing Course (Roberts Reference Roberts2014: 93–6), if we accept Roberts’s coding of cases, economic crises were deeper in the labour-mobilizing cases. But, as I argued above, the data in Table 5.1 attribute some country-years of economic performance to labour-mobilizing party systems even though military dictatorships that had crushed unions and parties were in power. Therefore, the evidence does not exactly show what it needs to. Party system instability was higher in the labour-mobilizing systems than in the elitist systems – again, if we accept Roberts’s coding of cases. This first condition shows correlation rather than causation, but it is a useful step.

Secondly, one should test the arguments with some control variables and/or alternative hypotheses. Without considering alternative hypotheses, it is difficult to assess arguments about complex phenomena. Changing Course does not seriously contemplate alternative hypotheses, nor does Roberts’s response to my review.

Changing Course invokes two causal mechanisms to explain why the labour-mobilizing cases experienced deeper economic crises, which in turn are a major part of the explanation for the greater party system instability (Roberts Reference Roberts2014: 93). However, the association between labour mobilizing party systems and deeper economic crises might be spurious. Two potential explanations for economic crisis that I mentioned briefly in my review merit systematic consideration, to which I will add a third.

1. (1) The economies that entered the era of neoliberal reforms with more adverse circumstances (e.g. higher debt burdens, higher inflation, worse fiscal deficits) experienced deeper crises during the transition period. Antecedent economic performance and conditions are a serious and measurable alternative explanation for subsequent economic performance, and they are exogenous to subsequent performance. Without controlling for them, it is impossible to know how much of the gap between labour-mobilizing and elitist party systems in subsequent years stems from dynamics generated by the party systems and how much stems from differences in beginning conditions.

2. (2) Bad policies could explain more serious economic crises. Roberts cogently argues that bad policies are sometimes endogenous to the kind of party system, but the reverse can also be true: bad economic results and severely adverse antecedent economic conditions can generate labour resistance.Footnote ² Changing Course does not take this possibility seriously enough.

3. (3) The countries that experienced more adverse swings in trade prices had worse performances during the neoliberal era. This hypothesis is relevant in a region of many commodity-exporting countries, and it is exogenous to the kind of party system. For example, the collapse in oil prices from 1980 to 1998 helps explain Venezuela’s bad economic performance during that time.Footnote ³

More importantly, Changing Course should consider alternative explanations for the magnitude of party system change. With 16 countries and by my count probably 73 electoral periods in Roberts’s data on electoral volatility (2014: Table 5.3), it would have been possible to do quantitative testing. Given the observational nature of the data, the limited number of observations and the impossibility of exhaustively testing for complex interactions among variables, quantitative testing could not have resolved questions of causality. But it would have been helpful in assessing the argument and some alternative explanations.

The argument that deeper economic turmoil was a primary cause of greater party system upheaval in labour-mobilizing cases (Roberts Reference Roberts2014: 52–3, 93, 96) makes deductive sense. However, the literature that addresses related topics does not support the idea that economic distress explains a great deal of the variance in party system upheaval. For example, studies of electoral volatility have consistently found that if economic performance has an impact, it is not substantively huge.Footnote ⁴ Among these works is an excellent article by Roberts and Wibbels (Reference Roberts and Wibbels1999) that covers from 1980 to 1997 for Latin America. Studies of party system collapse and party collapse have likewise found that even deep economic crisis does not come close to sufficing to explain massive party system change (Lupu Reference Lupu2016; Seawright Reference Seawright2012: 75–9). Therefore, the magnitude of economic crisis is not likely to explain as much of the variance in party system volatility as Roberts claims.

Thirdly, a crucial part of the evidence rests on showing the causal mechanisms at work: why labour-mobilizing cases generated deeper party system instability, and why elitist party systems caused less instability. Changing Course does not focus the case studies as a way of showing causal mechanisms. The case studies demonstrate deep knowledge of Latin America and keen analytical skills, but they do not systematically explore causal mechanisms. They (Roberts 2014: Chapters 7 and 8 on this point) do not carefully trace by what mechanisms elitist party systems generated more stability and labour-mobilizing party systems generated greater volatility during the transition era (roughly 1980–2000). The case studies in which outcomes diverge from the general theoretical arguments (e.g. Ecuador, an elitist party system with high volatility (Roberts Reference Roberts2014: 156–63); Chile, a labour-mobilizing party system with low volatility (Roberts Reference Roberts2014: 190–200)) likewise do not offer much detail about why the causal mechanisms did not work as expected.

The evidentiary requirements I have sketched here – coding cases so that the classification of cases matches the effects they supposedly bring about, considering alternative explanations, and tracing causal mechanisms – are standard practice in much of contemporary political science. They are demanding, but I do not believe that they are unreasonable. Changing Course articulates a clear, original and parsimonious hypothesis on a highly important issue, and it has many other strengths signalled in my original review, but in my judgement it falls short in testing the arguments. Practising these standards of presenting evidence well, of course, is a challenge to all of us.