Aggregate Data Analysis

Kinds of Aggregate Data

There are three types of aggregate data:

• 1.
  data relating to a collective as a whole (e.g., population or democracy indices, number of veto players or other characteristics of political systems);
• 2.
  aggregated individual data (e.g., the Gini Index for measuring income disparities, or the unemployment rate), which can be subdivided into two types:

  • a.
    census data, where every single case of the whole population is measured (complete inventory count)
  • b.
    data from samples, which should fit the context of the research question and be as representative as possible; and
• 3.
  event data, which give a frequency of events within a given period, mostly gathered from media sources (event data have become important sources for analysis, especially in international conflict research).

Basic Problems with Aggregate Data

There are some basic problems in comparative studies using aggregate data that have to be taken seriously. The first problem is whole-nation bias, which results from ignoring subnational variation; loss of information and reduced complexity are the consequences. Second, aggregate data are often measured in monetary units (e.g., for expenditure, import or trade volumes). For reasons of comparability between countries, the different currencies are most often translated into dollars or euros, which poses the problem of finding an [Page 54]adequate exchange rate for standardizing to the dollar or euro. One such standardization uses purchasing power parities, which may be misleading. Another alternative for comparing aggregated data is to standardize using a percentage of a national product measure. The choice and correct measurement of such indicators (e.g., gross domestic product [GDP] or gross national product [GNP]) are difficult and problematic; official data do not account for huge differences such as those concerning black markets or forms of subsistence income.

Intercountry comparability is also undermined by the use of different bases of calculation and a dissimilar inclusiveness of the indicators. An example makes this clear: The military budget is not identical with the budget of the Ministry of Defense. There may be a number of military expenditures (e.g., for the suppression of terrorism) that fall under the authority of other departments, such as the Ministry of the Interior. Another complication is the fact that the allocation of funds for such expenditure may vary from one country to another. The varying levels of quality in collecting data among countries can serve as a source of heteroskedasticity (i.e., a nonconstant variance where one expects a constant variance in the data). Therefore, different methods of data collection (e.g., estimation instead of more exact measurement) lead to less reliable data and, in the end, to poorer estimations in regression analysis.

...