In this post I argue that leaving culture out of consideration in empirical studies in the fields of Behavioral and Experimental Economics can put the generality of results found at danger. The starting point of my argument is the omitted variable bias. When relevant independent variables are left out in empirical analysis, the results obtained are most likely distorted. Relevant in this respect are those variables affecting the dependent variable other than the explanatory variables considered. Hence, the objective of the essay will be to show that culture is a relevant aspect to be considered.
While in theory it is easy to prove the distortive effect of omitted variables (cp. Wooldridge, 2002; p. 50), the main difficulty in practice is to identify what variables are relevant (and measure them). First, let us go one step back to look in greater detail at the consequences of the omitted variable bias and why its occurrence is so dramatic. Often we are interested in examining the effect of a certain exogenous shock. To carry out an empirical analysis we have to (a) develop a falsifiable model, (b) gather data fitting this model, and (c) test the model from (a) with the data from (b). As budget and time of researchers are limited, examining all individuals affected by the shock is not feasible. Therefore, a subset of the whole population that represents the entirety of the population is taken. From the analysis of the shock on this subset economists want to draw a conclusion on how the shock affects the whole population under consideration. However, if the sample diverges from the rest of the population in relevant aspects and this divergence is neglected in the analysis, a generalization will not be valid. [in essay example here]. Hence, the consequence of omitted variables is a model with limited generalizability. In the extreme case the relation found will only apply to the subsample.
Now, to see the severity of the omitted variable bias let us remember the very aim of economic analysis: economists want to reach general conclusions applicable to a great number of entities. These entities may take the form of different countries, economic systems, institutions, or others. From analyzing the effect of a shock on Country A we want to draw conclusions on how any other country would react to a similar shock. Hence, in this case Country A is the representative, the other countries are the entirety. From this we can conclude that differences in countries will lead to loss of generality. However, there are ways to deal with this.
In order to gain generality economists include a set of variables to control for distorting effects. We can identify two major features of variables that make them decisive for the generality of a model. A variable is decisive for the generality of a model if it
(a) has an effect on the dependent variable, and
(b) is different between the entities compared.
Note that (a) is the sufficient condition to include a variable in the analysis. If a variable affects the dependent variable but is not neglected, results will be distorted. The severity of this distortion depends on the size of the impact. If, however, this omitted variable takes the same value in all entities under consideration, generality is not harmed. (Note however that results will be distorted which is a fundamentally disturbing problem!) If now, additive to (a), the value of the omitted variable differs between the entities (b) the model found is not generally applicable.
There is a great number of variables accepted to fulfill the criteria named above. It is therefore common practice to include a battery of socio-economic variables such as gender, income, marital status, age, etc. Culture, however, plays until this day a minor role in the field of empirical economics. It is however nothing obvious that this is justified. Contrary to that, culture has to be considered whenever
(a) culture is assumed to have an effect on the dependent variable, and
(b) culture is different between the entities compared.
Again, while (a) is sufficient reason to include culture in the analysis in order to receive unbiased results, the combination of (a) and (b) limits the generality of results. Whether these two assumption do actually apply, critically depends on the characteristics of the study carried out – the dependent variable and the entities considered. The so far said serves as a theoretical foundation to analyze the role of culture. To facilitate the upcoming discussion we need, however, a common understanding of what we mean when we talk about culture.
While most of us have an intuition for what is meant by culture I deem it reasonable to state at this point the definition of culture I use in the essay. Cultural differences are defined by Roth et al. (1991) as “differences that cannot be attributed to variables other than the nation in which the data were gathered”. While this definition is an excellent starting point, a small adaption has to be made from my perspective. The notion by Roth et al. assumes that culture is homogenous within a nation. This is however nothing obvious, and even highly doubtable. It is questionable whether the borders of a country always coincide with the culture of the individuals living inside these borders. We may rather refer to the group of individuals sharing the same culture as a society. In this way we slightly adapt the notion of Roth et al. and define cultural differences as “differences that cannot be attributed to variables other than the society in which the data were gathered”.
The task left to the upcoming essay is to show in which cases culture has to be considered, which is to say, in which cases culture affects the dependent variable and differs between entities compared.
Roth, A. E., Prasnikar, V., Okuno-Fujiwara, M., & Zamir, S. (1991, December). Bargaining and Market Behavior in Jerusalem, Ljubljana, Pittsburgh, and Tokyo: AnExperimental Study. American Economic Review(5), pp. 1068-1095.
Wooldridge, J. M. (2002). Introductory Econometrics: A Modern Approach. Itps Thomson Learning.