“Neighbors in values”: A new dataset of cultural distances between countries based on individuals’ values, and its application to the study of global trade

Abstract

The paper proposes a method for evaluating the cultural distance between countries by analyzing the differences in values of individuals residing in these countries. These differences are assessed with an innovative approach based on MELNN ensemble of five metrics (Mahalanobis distance, Euclidean distance, L-distance, as well as normalized versions of Euclidean and L-distances), compiled through factor analysis. For each individual we find “neighbors in values”, i.e. other individuals with possibly closest MELNN scores. We proceed to build a network interaction model on the basis of MELNN scores. Analyzing the closeness to each individual from country A of the whole multitude of individuals residing in country B eventually allows to define the cultural distance between these countries. We use the thus obtained cultural distances in a gravity model, where they prove to be a significant factor influencing bilateral trade.

Introduction

The notion of cultural distance has received substantial attention in business studies on the companies’ foreign entry mode, cross-border acquisition, and foreign direct investment allocation (for a review see, e.g., Kandogan, 2012, Zinkina et al., 2016). The most popular and widely used approach to measuring cultural distance is based upon Geert Hofstede's cultural dimensions; not infrequently, researchers also use a simple standardized quantitative metric of cultural distance developed by Kogut and Singh (1988) on the basis of Hofstede's dimensions (for a review of such papers see, e.g., Kirkman et al., 2006, Samiee, 2013).

However, in 2001 this approach came under criticism by Oded Shenkar, who emphasized a number of inherent drawbacks, such as the illusions of symmetry, stability, linearity, corporate homogeneity etc. (Shenkar 2001).

Gradually, researchers started turning their attention to other possible ways of constructing the measurements of cultural distance. Some works (e.g., Drogendijk and Slangen 2006) used the theory of basic human values by Schwartz with its 7 dimensions.

Recently, a number of papers aimed at measuring cultural distance took World Values Survey data as their basis. Thus, a paper by Tadesse and White (2010) calculates the cultural distance between each of the nine OECD reference countries and each of the remaining countries in their 67-country sample using two dimensions of culture developed by Inglehart: traditional vs. secular-rational authority (TSR) and survival vs. self-expression values (SSE). They use Euclidean distance to define cultural distance between countries I and J, which is calculated as a square root from the sum (TSRj-TSRi)² + (SSEj-SSEi)².

Cyrus (2012) uses responses to WVS questions regarding trust, respect, control, and obedience to define Euclidean cultural distance between countries A and B, which is measured as the square root taken from the sum of squares of differences between trust, respect, control, and obedience scores for countries A and B.

Kaasa et al., (2016) create a set of cultural distance measures both at the country and regional levels in Europe which are calculated using Hofstede dimensions and the data from EVS and WVS. The measures are presented in the form of matrices.

Another approach to using WVS in measuring cultural distance is notable for using the micro-level data to build clusters of respondents who are similar to each other within clusters and differ between the clusters. Cultural distance between countries is small when their citizens are distributed among the clusters in a similar way. This distance can be expressed through the distance between the observed distributions (the average of the squared differences between the ranks) (De Santis et al., 2014).

In this paper we propose a different approach to measuring cultural distance. In line with the approaches mentioned above, it is built on analyzing the differences in cultural values between the populations of different countries. However, our approach is innovative in a number of aspects. We base our measures not on the aggregate (country) level, but rather on the level of individual respondents. The former approach does not take into account the within-country dispersion of answers. Going down to the level of microdata (the answers of individual respondents) allows to overcome this drawback. We ensemble 5 different metrics to find the “neighbors in values” for each individual respondent. After that, in order to define the cultural distance between countries, for all individuals belonging to a given country we analyze the structure of their “neighbors in values”. As we define an individual network model we can also use in our analysis not only the direct neighbors, but also “neighbors of neighbors” etc. This allows constructing the metrics to assess the distance between countries. Moreover, this approach takes into account the principal non-linearity which arises at quantitative estimation of such complex and multilateral phenomenon as “human values”.

The paper is structured as follows. The introduction presents a brief review of papers aiming to measure cultural distance on various data, as well as a short description of our own approach. In Section 2 we describe our procedure of choosing variables. Section 3 presents an ensemble of metrics we use to analyze the data and to obtain individual scores (how far each respondent is from each and every other respondent). Section 4 explains how we use these scores to set a network model, and proceed to use this model to define each individual's “neighbors in values”. It also explains how we differentiate between “neighbors” of different orders. Here we also display the algorithm of transforming the individual scores of closeness into the country scores of closeness. Section 5 displays the results: here we present the closest and the most distant “neighbors in values” for some countries with their respective scores of cultural distance (for the full set of cultural distances between each and every country in our sample see Appendix A). In Section 6 we test the robustness of our cultural distances by using them in the gravity models with various specifications. For the sake of comparison here we also re-test the approach by Tadesse and White (2010) on more recent data. Section 7 presents the results of gravity model estimates and concludes, bringing in some discussion points, such as the perspectives of further usage of our measures of cultural distance.