Measurement, Scales

In the natural sciences, scaling is usually very precise, and a single measure is often sufficient to describe a dimension—for example, when measuring temperature in Celsius, weight in grams, or [Page 1528]distances in meters. In these cases, the assignment is quite simple: Distances can be measured by a ruler, temperature by a thermometer, and weight by a scale. The situation in economics is somewhat different. One of the most important indicators in international finance is the purchasing-power parity. The validity of this measure depends on the choice of an appropriate “basket” of consumer goods, which is not the same in all countries. In 1986, The Economist magazine introduced “Burgernomics” and has been publishing since then annually the prices of McDonald's Big Mac hamburgers in different countries. The advantage of the Big Mac is that it is produced and consumed around the world; the unique basket consists of prices for beef, cheese, lettuce, onions, and bread—and since the recipe is the same worldwide, there are (almost) no differences in the share of these five ingredients among countries. Note that the one-dimensional Big Mac Index consists of several indicators that have different weights—there is, for example, a higher share of beef than onions in a hamburger.

In the social sciences, scaling is the process of measuring or ordering entities with respect to quantitative attributes. While natural scientists use instruments that are as precise as possible, social and political scientists use survey research methods such as face-to-face or telephone interviews to evaluate the attitudes and beliefs of the population. In the following, some examples from large data sets such as the International Social Survey Programme (ISSP), the World Value Surveys (WVS), and the American Nationwide Election Studies (ANES) are shown to illustrate problems of frequently used measurement scales such as Likert or Guttman scales.

Typical research questions include the discussion of latent variables (dimensions, factors) such as the level of national identity or the level of political efficacy and trust. The validity of the measures depends on the quality of the instruments: It is not possible to measure the correct length with a broken ruler, and it is not possible to measure true attitudes and beliefs with poorly worded questions. Starting with a single question, closest to a metric variable might be a feeling barometer that runs from 0 to 100. The ANES 2004 provides the following question: “Where on that thermometer would you rate George W. Bush?” The problem with this kind of question is that respondents are not able to differentiate between values such as 43 and 44; most are not even able to differentiate between scores such as 62 and 76. An alternative is to reduce the scale to values between 0 and 10 or, for getting an easy-to-catch midpoint, to values between –5 and 5, as is done, for example, in the German Politbarometer. Although it is probably easier to respond to an 11-point scale than to a 101-point scale, it is still difficult to differentiate between categories such as 2 and 3.

...