Observing Unexpected Patterns in Cross-National Research: Blame Data, Theory, or Both? Attitudes toward Redistributive Taxation in Thirty-Three Countries

Abstract

This article examines the relationships between socioeconomic status and attitudes toward redistributive taxation across 33 countries using the complete International Social Survey Programme (ISSP) 2006 data set. We apply a simple rational-choice-inspired homo-economicus hypothesis proposing that those better off in the socioeconomic hierarchy should have less reason to support state-organized economic redistribution compared to those situated at lower levels in the socioeconomic hierarchy. The empirical results demonstrate substantial cross-country variation regarding the correspondence between empirical observations and theoretical expectations. When faced with such tremendous cross-national variation in response patterns, a common strategy among researchers is to question the quality of the data collection procedures for those countries deviating strongly from theoretical expectations. The strategy chosen in this study is, however, different. The main argument is that an observed lack of fit between theory and empirical observations may be rooted in problems related to theory rather than the quality of data collection procedures. Building on previous research, two “cultural distance” hypotheses are formulated, both of which argue that the correspondence between the homo-economicus theory and empirics should indeed vary systematically across countries. The first focuses on the role of the welfare state and the second on the level of economic affluence and associated scientific dominance. Both hypotheses receive considerable empirical support. The relationship between socioeconomic status and support for redistributive taxation is substantially stronger in the wealthy Western welfare states—particularly among those of Northern Europe—than in the poor non-Western countries lacking any institutional design reminiscent of a welfare state.

Keywords:

• attitudes
• international comparative research
• ISSP
• latent class analysis
• taxation

ACKNOWLEDGMENTS

The authors are grateful to Christof Wolf and Markus Quandt for commenting on a previous version of this article. In addition, the authors thank the editors and the external reviewer for very valuable concrete comments during the finalization of the article.

Notes

An example of an empirically generated theory based on a translation mistake was recently reported by Charles Kurzman in the Washington Post (Kurzman 2014; for the translation check procedure, see Mellon 2011).

Having said that, we will of course not completely ignore focusing on measurement errors related to data. However, we will restrict our attention to the countries displaying the most aberrant response patterns.

The attitudinal items used in this study have also been constructed through a Western lens. They actually appeared in the first ISSP survey back in 1985, many years before countries from the non-Western world joined the program.

For a broad overview of methodologically related issues in comparative research, see Harkness et al. (2010a).

This section builds on arguments and empirical applications provided by Edlund (2007), Edlund and Sevä (2013), and Edlund and Lindh (2015).

For detailed information on national sampling procedures, data collection, and response rates see the ISSP 2006 Study Monitoring Report (Scholz et al. 2008).

The latent class analysis software used in this study is LatentGold 4.5.

Please note that the ISCO-based component of the index is a rather rough measure in the case of South Africa and Venezuela, since the ISCO coding is available only as a one-digit code.

The total number of valid responses for the tax items is 42,162. Out of these, 40,125 have a valid response on the socioeconomic index.

One caveat, however, is that data covering all 33 countries were available only from 2009. However, for the countries where time-series data are available, an analysis shows that the levels of government spending and taxation tend to be rather stable between 2006 and 2009.

ISSP Working Principles: “non-literal translations, that is, ‘culturally equivalent questions,’ are to be notified to the group and indicated in the documentation.” Apart from that there are no mandatory rules for the procedure of the translation process.

The languages covered by the Philippine survey are English, Tagalog, Ilocano, Bicolano, Ilonggo, and Cebuano. In South Africa the languages are English, Afrikaans, Venda, Xhosa, Zulu, and Tsonga. According to the ISSP Study Monitoring Report (Scholz et al. 2008), all translations were carried out by trained translators and checked by language experts.

Here we would like to thank Ivet Solanes Ros and Webster Whande for their support.

Thanks to Gerado A. Sandoval from the Social Weather Stations in the Philippines.

The poverty and subsistence thresholds, respectively, are absolute figures, comprising people who cannot afford to buy a national specific minimum basket of goods. The subsistence threshold lies even below the poverty line and indicates the very minimum level of what a person needs to survive. The definitions of these levels are set by the countries.

We are aware that there are additional dimensions of poverty that cannot be comprehended by measuring pure income levels. Furthermore, there exist different income-based measures of poverty as well as poverty lines (see, e.g., Ravallion 2010). Apart from that, in urban economies with large informal sectors, income flows may be erratic. The same is true for agrarian economies where household incomes rely on the harvest cycle (Coudouel, Hentschel, and Wodon 2002). Nevertheless, the comparison of income statistics with the income information in the survey data should be sufficient to figure out whether an entire group of the society might be excluded from the survey.

Incomes tend to be underestimated in surveys, especially with only one income question, because respondents simply forget to include all income sources (DAE 1998). However, that does not account for those respondents living at the existence minimum. Those who have very little, are usually able to keep track of it. Since the age structure of the household members is unknown, the following formula was used: equivalence income = family income / (household size * 0.73). (For further information regarding the construction of equivalence income see DAE [1998].)

National Statistical Coordination Board, Philippines, 2006.

The Presidency of South Africa (2008) Development Indicators 2008: Van der Berg (2007) based on AMPS of various years (1993–2007), Bhorat (2007) based on Statistics SA IES data (1995, 2000, and 2005), and Statistics South Africa and National Treasury (2007:23). ZAR stands for South African Rand.

Further analyses have been done taking into account different regions and therewith different regional subsistence lines for the Philippines and South Africa. The results do not change the cross-regional data outcomes.

Additional information

Notes on contributors

Insa Bechert

Insa Bechert is a researcher at the GESIS Leibniz Institute for the Social Sciences in Cologne, Germany. Her main research field is the comparability of measurement instruments in international comparative sociology. She is responsible for the International Social Survey Programme (ISSP) data in the GESIS Data Archive for the Social Sciences.

Jonas Edlund

Jonas Edlund is a professor of sociology at Umeå University in Sweden. His main areas of interest are comparative political sociology and stratification. He heads the Swedish part of the ISSP.