Choice of Variables

In the following paragraph, the variables included in the linear regression model will be specified and reasons for the choice of our model will be given.

We intend to identify the major drivers of research intensity for the countries appearing in our data.

A country with high research intensity is often targeted in research. In comparison, a country with low research intensity receives low attention in research and is expected to appear less often in our dataset. Therefore, a statistical model is required that can predict research intensity about countries based on the major driver of such a value. To measure research intensity, the number of times one country appears as single-country study during the whole 15 years across all strategy and management journals in our dataset will be chosen as the dependent response variable. Deng and Zhu (2015), Chan et al. (2007) and Das et al. (2013) tested a model that identified a strong research-wealth relationship in their respective dataset. As predictors, they tested for the significance of GDP per capita and the population size of the included countries. Therefore, we hypothesize that those are major drivers of research intensity that are expected to explain a high proportion of variability in our dependent variable. In line with Das et al. (2013), average values for those two predictors are taken for the period of 15 years. Data for the respective average GDP per capita and the average population of each country are taken from the World Development Indicators, published by The World Bank (2017).

Hong Kong and Macao are treated separately from China. Data for Taiwan are supplemented by the statistics of the Republic of China (Taiwan) (2018).

To further understand what country differences might influence the decision of authors when choosing a specific target country, distance measures were identified that particularly account for

the concept of psychic distance seemed to be interesting. Originally, psychic distance is defined as the subjectively perceived distance to a given foreign country. As Håkanson & Ambos (2010) explain, perceived distance to a foreign environment is commonly associated with higher difficulty of collecting, analyzing or interpreting information about it. This concept is often targeted in international business studies and the context of establishing businesses abroad. However, we hypothesize that psychic distance also leads to difficulties of collecting, analyzing and interpreting information for any researcher. To meet quality standards in academic publishing, access to information, correct analysis and interpretation of the findings are key. Therefore, we expect that higher psychic distance to the US is associated with low research intensity about such a distant and foreign country, from the perspective of US affiliated authors. The US is chosen as base country due to the primary US origin of the chosen journals.

There is no known single measure for psychic distance, although it has often been approximated by cultural distance effects alone. Around 1970, Professor Geert Hofstede conducted one of the most comprehensive studies of how culture influences values in the workplace. Based on around 116.000 survey questionnaires given to IBM employees in 72 countries, four national culture dimensions were identified. The study was later extended to a wider range of countries and two further dimensions were added to the original model. The concept of national culture only applies to nations as a whole and the application on individual levels therefore receive plausible critique (Minkov & Hofstede, 2011). US affiliated authors do not represent the whole average American culture, however their perceived distance to other countries refers to other countries as whole nations. Therefore, it is believed to be reasonable to apply Hofstede’s model in spite of some limitations. Based on Geert Hofstede’s cultural dimensions, Kogut and Singh (1988) invented a single measure for cultural distance, that combines the four original cultural dimensions, known as power distance, individualism, masculinity and uncertainty avoidance to one single index. This index has received criticism especially for the ignorance of two later added cultural dimensions and for giving each cultural dimension equal importance (Håkanson & Ambos, 2010). Nevertheless, the index will be included in our model, since it has not been tested in the underlying context. Notably, there are several proxies available to capture cultural distance besides the dimensions of Hofstede, namely Hofestede’s cultural dimension, Schwartz value survey and the Globe study (House & Global Leadership and Organizational Behavior Effectiveness Research Program, 2004; Schwartz, 1994). However, it is questioned whether any measure between those three is superior to another and correlation between the three variables is not as high as one might think (Beugelsdijk, Ambos, & Nell, 2018). Therefore, those three variables might hardly be comparable. Since the variable is chosen as proxy for cultural

distance, it is of essential importance to report which measure was chosen. Instead of trying to balance reasons for or against one of the available proxies, which has led to inconclusive findings of different authors in the past, Hofstedes’ four cultural dimensions were chosen, based on the original dataset.

The two further dimensions were not included since they were not produced in the same original study. To our knowledge, this is the first time such regression model is applied to the underlying context. Therefore, the application of Hofstede’s for cultural dimensions through the Kogut & Singh index serves as first step in the development of such a regression model, since it is a widely known measure that readers might most likely be able to relate to. As cultural distance is intended to be tested as sole variable, this index serves the requirement of one single measure for it and dimensions will not be tested separately. Calculations of the respective four values for each country in our data and the derived cultural distance to the US are based on the data provided by Hofstede (2010).

The index formula takes the form of RS_T = ∑^__+`'VWX_+T− X_+YZ^:/\₊] /4,

“(…) where X_+T stands for the index for the -_abcultural dimension and c_ab country, \₊ is the variance of the index of the -_ab dimension, d indicates the United States, and RS_T is the cultural difference of the c_ab country to the United States (Kogut & Singh, 1988, p. 422).”

Additionally, we test for the significance of geographic distance to the US. In their assessments of antecedents of psychic distance, Håkanson and Ambos (2010) found absolute geographic distance to be the major driver of psychic distance perceptions. We rely on similar data for geographic distance and use data provided by the CEPII database (2018). Of several options available to measure geographic distance between country pairs, the distance between capital cities is applied. This is arguably not the best choice, since capital cities are not necessarily situated in the geographic center of a country. However, capital cities hold a crucial role as cultural, economic, political and social centers of a country and therefore can be seen as specific country centers applied for distance measures.

Finally, we add a dummy variable for English as an official language to our model. Psychic distance perceived is negatively associated with a shared common language between an individual and a foreign environment. The smaller the psychic distance to a foreign country, the more we expect this country to be a research target by US affiliated authors. Therefore, we assign the value of 1 to countries where English is an official language and the value of 0 to countries with an official

language different to English. Data for this variable are retrieved from The World Factbook by the Central Intelligence Agency (2018).

Therefore, besides including the research-wealth relationship in the model, the significance of a widely used approximation index for psychic distance is tested and additional more modern variables that have shown to be of high importance in capturing psychic distance effects are added as well.

4 Results

The results section seeks to answer the two initially formulated parts of the broad research question and their respective subordinate questions and hypotheses. Therefore, this chapter is also divided into two main sub-sections – the descriptive results of the bibliometric study and the linear regression results.