Family Systems and the Historical Roots of Global Gaps in Democracy

ABSTRACT

The current study investigates the role of ‘family systems’ as a historical institution in explaining why some countries have enduring democracy while others remained authoritarian despite the repeated global waves of democratization. To do so, empirical data including information on 127 countries between 1849 and 2009 has been gathered. The results of cross sectional and panel data analyses show that countries characterized by a nuclear household structure in the past also have higher levels of democracy in the long run (at the national level). Thus, the current study provides evidence for Todd's hypothesis on the origins of political systems. Moreover, family systems that determine the position of women are also found to be relevant for democratic development. The persistent effect of family systems on democracy can be attributed to their link with norms and values that are conducive to democracy, gender equality and local democracy practices. Overall, these findings emphasize family organization as an important historical factor in understanding the long-term global patterns of democratic development.

Keywords:

• family systems
• historical institutions
• democratic development

JEL Codes:

• J11
• Z10
• Z13
• D72
• N30

Notes

2 See map at Figure A1 in the Appendix for a global illustration of the political regimes. This map is based on the Polity IV index, for which information is provided below.

3 According to proponents of this view, countries democratize in clusters of geographical regions. For instance, the third wave includes the democratic transition of Southern European countries in the 1970s, Latin American countries in the 1980s, and Soviet bloc countries in the 1990s (Wejnert 2005). Lipset (1959) described this as the snowballing effect. The diffusional pattern suggested that the likelihood of a transition in a country depends on the events in other states (Wejnert 2005).

4 In China, civil liberties are argued to have improved significantly with three decades of sustained economic growth; however, an evaluation of this claim is beyond the scope of this study (Ginsburg 2007).

5 While a number of studies have suggested presence of a theoretical link between the family organization and democratic development, the empirical evidence on this link between family systems and democratic development is scarce. An exception to this is the study by Lane and Ersson (2005), which considers the relevance of family values for democracy. However, the authors use the family classification by Todd (1985) only to rank countries on a individualism scale and their analysis is limited to the recent time period. A similar study on the implications of family systems for democratic development through norms and values has been provided by Eckstein (1966), focusing on Norway as a case study.

6 As an alternative to Todd's model, Therborn (2004) offers a geo-cultural definition of the family systems on a global level; however, his model does not highlight within-regional differences, and therefore is not ideal for cross-national level analysis.

7 The exception to the statement about the regional variation is Africa where variation is present only in the northern part and Southern tip of the continent (please see Figure A2 in the Appendix).

8 Dahl (1971) uses the term ‘polyarchy' rather than democracy because for him, democracy is an unachievable ideal type (Fish 2002).

9 Here the sample was limited to the time period from 1849 onwards, as the data coverage for the variables included in the analysis become substantially better starting from the second half of the nineteenth century.

10 Universal suffrage is not one of the dimensions covered in the Polity IV index, for which it has been commonly criticized (e.g., Bollen 2009, Paxton 2000). To address this issue, the Freedom House index was employed as an alternative indicator of democracy.

11 A continuous measure of democracy is preferable over a dichotomous measure of democracy, which groups countries as being democratic or not. In a dichotomous measure, a large share of the variation both within and between countries is lost (Bollen 2009).

12 The correlation of the Polity IV index with other democracy indices in the literature created by Bollen (1980), Arat (1991), Vanhanen (1990) and Gasiorowski (1993), ranges between .85 and .93 (Jagger & Gurr 1995).

13 The standardization of the scales has been done by dividing each index by their maximum value. The Polity IV index was first recoded it to range between 0 and 20 and then diving it by 20.

14 The measures on the quality of governmental institutions have been combined into a single dataset by Teorell et al. (2013), which is used in the current study. More information on the data can be found in Table A1 in the Appendix.

15 For a detailed description of the democratization waves that took place since the early nineteenth century, see Huntington (1991).

16 An exception to this rule discussed by the authors is Argentina.

17 These values include social toleration, life satisfaction, public expression and an aspiration to liberty.

18 There are contradictory views on whether a tension exists between the two principles of democracy, liberty and equality. However, this discussion is beyond the scope of current study. Please see Lom 1999 for a discussion on this topic.

19 As devised by Hajnal (1965).

20 Todd's family systems have also been shown to matter for pension systems (Galasso & Profeta 2010), intimate-partner violence (Tur-Prats 2014), fertility (Rotering 2014), human capital formation (van der Vleuten 2014), and gender equality (Dilli et al. (2014)).

21 The differences between the family systems on intermediary channels presented in Table 3 are also statistically significant, taking into account the same set of control variables described in section 4. As these results are not the main interest of this study and therefore excluded, these results are available upon request. Moreover, majority of the variables concerning the intermediary channels are available only at one point in time (mostly current day data, except for structural characteristics), limiting possibilities for a structural equation modelling.

22 The figures on local democracy measure might be driven by the fact that Murdock's Ethnographic Atlas has better coverage for Africa compared to other regions of the world (Rijpma and Carmichael 2013).

23 See Tur-Prats (2014) for a discussion.

24 It is not possible to include country fixed effects because of the time invariant nature of family systems.

25 Inclusion of measures on income inequality and industrialization do not change the interpretation of the findings. As the number of observations on income inequality is relatively smaller compared to the rest of the indicators in the model and industrialization measure is highly correlated with log GDP per capita, they were excluded from the models presented here.

26 Religion is also included as percentage of the population that have Muslim, Protestant or Catholic religious denomination in 1900, 1970 and 2000, and conclusions that could be drawn from this specification remains the same.

27 These regional and time fixed effects also capture the existence of ‘diffusional pattern' and ‘democratic waves’.

28 The endogeneity issue can also arise due to reverse causality and the measurement error and. The reverse causality issue is discussed in the previous section. While this issue of reverse causality cannot be eliminated completely, based on the secondary literature, it is more plausible to expect a relationship running from family systems to democratic development rather than the other way around. Moreover, as the family systems measure is only available at the national level, within-country variation could be a source of a measurement error. However, as the main dependent variable, democracy is captured at the national level, this is not a major concern. Another issue is the time invariant measure of family systems. Table 5 presents the results when family systems are assumed to change due to economic development or other drivers. Under a set of different assumptions, family systems remained as a robust predictor of democracy.

29 Two issues make it particularly difficult to find an instrument in this study. The first reason is mainly due to the suggested historical relation between family systems and state development, therefore any instrument is likely to be correlated with our democracy measure as well. More insight is needed on the origins on family systems to be able to find a relevant instrument. A second issue concerns finding an instrument that would be valid for each seven categories of family systems that have different characteristics from each other.

30 An exception to this rule is arguably France that has an egalitarian nuclear family system and one of the longest modern democratic traditions in Europe. The ideology of popular participation in political decision-making stretches back to the Revolution of 1789 in France. There was, for example, the right to vote for the vast majority of adult males during the revolutionary years of the 1790s. However, while being the host to the French Revolution of 1789, arguably, a stable democratic regime did not emerge in France until after the Second World War, with the establishment of the Fifth Republic (Heywood 2004, 921).

31 Inclusion of these indicators on development and historical characteristics in Model 2 and Model 3 also do not lead to a substantial change in R2 (see Table 4).

32 All the measures on family systems have been employed from this dataset (See Table A1 in the Appendix). Rijpma and Carmichael (2013) use not only the family systems classification by Todd but also employ data from Murdock's (1959) Ethnographic Atlas to check the validity of Todd's classification. Bolt (2012) has aggregated the measures at the ethnic level from Murdock dataset to the country level. For a detailed description, see Rijpma and Carmichael (2013) and Table A1.

33 Early marriage is a dichotomous variable in which 1 refers to Singulate Mean Age of Marriage below 23 and 0 above 23. This threshold is determined by taking the European Marriage Pattern as a reference point. The rest of the indicators are also captured in terms of dummy variables with 1 referring to family practices that discriminate against women.

34 The principal aim here is to test the model of Todd (1985); therefore, the variable on asymmetrical inheritance practices is kept in the analysis rather than the patriarchal inheritance variable.