Skip to main content
SpringerLink
Log in

A Measure of the Cultural Distance Between Countries

  • Published:
Social Indicators Research Aims and scope Submit manuscript

Abstract

We present a new method for evaluating the relative distance between any two countries, among several, using individual data. We form clusters of respondents and we calculate the proportions of each country’s respondents who belong to the various clusters. Assuming that respondents in the same cluster are similar to one another and that two countries are close to each other when their nationals distribute similarly among clusters, the dissimilarity between countries can be expressed in terms of Euclidean distances between the observed distributions (the square root of the sum of the squared differences between the proportions of nationals in the same cluster). We test the method on the World Value Survey (WVS) dataset for the years 1994–2007, first separately, by “domain” (opinions and attitudes on, e.g., religion, politics, and family), and then on all of the (selected) variables together. Groups of assumedly similar countries (the Baltic, the Nordic and the Mediterranean countries) turn out to be closer to each other than do, on average, any two European countries picked at random, which lends credibility to the method. Its pros and cons are discussed in the final section.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

Similar content being viewed by others

Notes

  1. In this paper we speak of indicators (or manifest variables) to denote variables that can be observed (i.e. the answers to the questions of the VWS—see Sect. 2.4) and domains (or latent variables, or dimensions) to denote variables that cannot be directly observed or measured, but that we assume exist and influence the corresponding indicators. For instance “religious attitude” is a latent (not directly observable) variable (or domain, or dimension), about which something can be said by observing a few empirical indicators (answers to such questions as “How often do you go to mess?” or “Is religion important in the education of a child?”).

  2. We also tried a few alternative specifications, introducing some arbitrary metrics. The results, not shown here, are in line with (actually, even slightly better than) those that will be presented shorty. All the elaborations were run with the SAS software, release 9.3.

  3. In the survey these were two separate questions, on the first and the second most important aim (questions E001 and E002). Since Jaccard multiplies the possible combinations of cases, which rapidly exceeds computing capacity, we had to circumvent this limitation in various ways: in this case we merged the answers to the two separate questions, ignoring the difference between the first and the second most important aim of the country, in the respondent’s opinion.

  4. Take, for instance, the domain “Gender attitude” (Sect. 2.4). It has five corresponding manifest variables, the first with 3 possible answers, the others with 4. This makes 3 × 44 = 768 theoretical different typologies (some of which empty, to be sure), and a corresponding matrix of 7682 = 589,824 (theoretical) distances, which is very large, but still manageable.

  5. A standardized variable z can be obtained from an original variable x with the following transformation z = (x − A)/σ, where A is the average of x, and σ its standard deviation. By construction, a standardized variable z has average A z  = 0 and standard deviation σ z  = 1. Ideally, with results independent of the number of clusters, one should obtain (roughly) straight lines z for each group.

  6. With a different number of clusters results were very similar, as Fig. 2 suggests (not shown here).

  7. The scores have more or less the same range on all the manifest variables. But since we dealt with manifest variables the possible answers to which ranged between 2 -dichotomous- and 21 we adjusted the scores so as to catch their contribution to that latent domain to the best of our possibilities. With alternative scores, not qualitatively different to be sure, the results did not change appreciably (not shown here). For the details, please refer to (De Santis et al. 2014).

References

Download references

Acknowledgments

We thank two anonymous referees and our colleague, Leonardo Grilli, for their useful suggestions.

Author information

Authors and Affiliations

  1. DiSIA - Department of Statistics, Informatics, Application “G. Parenti”, University of Florence, Viale Morgagni 59, 50134, Florence, Italy

    Gustavo De Santis, Mauro Maltagliati & Silvana Salvini

Authors
  1. Gustavo De Santis
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Mauro Maltagliati
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Silvana Salvini
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Gustavo De Santis.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

De Santis, G., Maltagliati, M. & Salvini, S. A Measure of the Cultural Distance Between Countries. Soc Indic Res 126, 1065–1087 (2016). https://doi.org/10.1007/s11205-015-0932-7

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11205-015-0932-7

Keywords

Search

Navigation