The Persistence of Inter-war Cleavages: Agrarian Inequality and the Left
By Jordi Domènech and Ignacio Sánchez-Cuenca
Lipset and Rokkan famously claimed that the modern structure of cleavages was created in the interwar period, with the generalization of universal male suffrage and the advent of mass politics. They tried to explain the stability of Western party systems in the postwar era with the hypothesis that the interwar cleavages somehow became “frozen”. The “freezing” hypothesis has been the subject of much analysis and debate. We suggest that this rich comparative literature can be approached from the viewpoint of historical legacies, where the effect of an extinct cause survives vast economic, social and political transformations. After all, “freezing” is an informal way of referring to persistence.
In previous work, one of us found an intriguing association in affluent countries between the patterns of political development in the interwar years and the wave of political radicalism and violence in the period going from the late 1960s to the 1990s. The countries that suffered more instability in the interwar period were also the ones in which the wave of radical protest became violent. This was interpreted in terms of persistence of certain forms of political radicalism that were forged when the Left was repressed by the reactionary autocracies that followed the democratic breakdowns of the 1920s and 1930s.
In a more recent study, we have used the historical legacies framework to analyze a cleavage that did not get much attention from Lipset & Rokkan, namely the agrarian one, grounded on access to land. Based on the work of Stefano Bartolini in The Political Mobilization of the European Left, 1860–1980 (2000), we consider that, in late industrializing countries, the agrarian conflict was more central than the industrial one in the interwar period. In addition, among late industrializers, the agrarian issue was more divisive the higher the level of agrarian inequality: typically, distributive conflicts about land ownership reached its maximum in areas dominated by latifundia, with a few large landowners and a vast number of landless peasants. Our main hypothesis is that, in the latifundium regions of the industrial laggards, peasants leaned to the Left and adopted more radical positions. We also argue that these political preferences have survived until today. In early industrializers, in contrast, land inequality was not a relevant factor in the formation of political cleavages, especially because only a minority of workers were employed in the agrarian sector when universal suffrage was established.
We posit there are two main mechanisms to account for the persistence of Leftist preferences, one economic and one political. The economic channel captures the negative impact of agrarian inequality on long-term development, typically in the form of higher poverty and lower human capital. The political mechanism is based on the generational transmission of political identities and loyalties. These political preferences first crystallized in the interwar years.
In order to test our main conjecture, we focus on the case of Spain (an industrial laggard with high land inequality) and we briefly compare the results with those of Italy (a similar case to Spain) and Britain (an early industrializer with high land inequality). According to the argument, Italy should behave like Spain, whereas an electoral legacy of land inequality should not be apparent in the case of Britain.
Spain is an ideal case for several reasons. Firstly: it had a very heterogenous structure of land ownership, with regions dominated by large landholdings and regions with family farms. Secondly, political preferences crystalized during the Second Republic (1931-36), a democratic period in which land reform was perhaps the most salient issue. Thirdly, the onset of Civil War and the Rightists’ victory meant the ambitious plans of the 1932 land reform were ditched. And lastly, agrarian conflicts were of secondary importance when democracy returned in 1977.
Electoral results at the provincial and municipal level covering the whole democratic period 1977-2019 (15 general elections), as well as preferences measured by surveys, are our main outcome variable. We measure agrarian inequality in various ways and using different sources of information (like Spain’s 1860 and 1920 population censuses and cadastral information from the 1920s). No matter what indicator we use, we obtain a very robust effect of land inequality in the period 1860-1930 on support for the Left in 1977-2019. The following table shows the baseline results. Column (1) presents the provincial analysis; column (2) the municipal analysis for a sample of 800 municipalities in which agrarian inequality can be estimated in the 1920s; and in column (3) an analysis of a large survey conducted immediately before the first democratic elections in 1977. In every case, the effect is positive, substantial, and statistically significant.
We use instrumental variables to enhance the causal interpretation of the main effect. Especially, we deal with measurement error in our main explanatory variable by instrumenting the various proxies of agrarian inequality at the provincial level with a measure of the speed of the Reconquest. The Reconquest is a period of almost eight centuries in which the Christian kingdoms of the north of the Iberian peninsula moved South conquering the various Muslim kingdoms. In the first part of the Reconquest, the Northern part of Spain was re-populated with free peasants who owned land. It was the last 400 years of the Reconquest that were connected to acute agrarian inequality. In contested, very depopulated areas, the aristocracy and the military religious orders protected the peasants and owned most of the land. Using the Speed of Reconquest as an instrument for historical agrarian inequality, our estimates of the effect of past agrarian inequality on Leftist voting are larger in the two stage least squares regressions using all the elections or each individual election.
In order to understand the main mechanism at work, we have analyzed the role of several mediating variables. Our main interest is the political channel, which we surmised operated through the formation of political cleavages in the 1930s. To proxy the latter, we have collected data on voting for the coalition of Leftist parties (the Popular Front) in the February 1936 election and the level of Rightist repression levels during the Civil War.
We also consider an economic channel, capturing the deleterious effects of agrarian inequality on long term development (we proxy these intermediate effects using contemporaneous unemployment, industrialization and educational levels). We perform sequential g-estimation to correct for potential post-treatment bias, following Acharya, Blackwell and Sen (2016). The main conclusion from these exercises is that most of the effect of land inequality goes through the political channel; the economic channel also plays a role, but of lesser importance.
Given the characteristics of Franco’s dictatorship (banning of all leftist organizations, a Church aligned with the regime, education under the grip of the state), we posit that the most likely mechanism of transmission of political preferences was the family. Using an exceptionally detailed survey of 2008, with questions about the ideology of parents and the side of the family during the Civil War, we find that in those provinces with higher land inequality respondents attribute more leftist positions to their parents (controlling for a battery of individual and provincial variables). As expected by the theory, the effect is reduced once we control for the side the family took during the Civil War (a relevant intermediate variable according to the mediation analysis). The basic results can be seen in the following coefficient plot.

Lastly, we address the external validity of our hypothesis. According to it, we should find a similar effect in Italy and a null one in Britain. The data confirm these expectations. In the case of Italy, we correlate the percentage of landless peasants in 1931 and support for the Left (and specifically for the Communist Party) in the 1970s. The effect is comparable to that of Spain (a higher percentage of peasants is associated with higher support for the Left).

In England, industrialization predated democratization. Hence, the labor-capital conflict should be more relevant for the formation of cleavages during the interwar period than the agrarian one. Using several indicators of agrarian inequality and poverty from the 1830s and 1873 for English counties, we find no association with support for the Labour Party in the 1970s. As the next figure shows, the relationship is negative (and insignificant).
In sum, we conclude that, in an industrial laggard such as Spain, agrarian inequality had long lasting political and economic effects, but the political ones are more important to predict electoral outcomes in the contemporary democratic period. In the interwar years, landless peasants developed leftist preferences that persisted through intergenerational family transmission and, to a lesser extent, through economic backwardness and late development.
These results suggest that the literature on cleavage formation and cleavage freezing is worth revisiting from the perspective of contemporary studies on historical legacies and persistent effects. We need a more systematic analysis of how political cleavages were formed in the interwar period and the mechanisms that have contributed to their persistence or decay.