(Page 9 of 14) - Scope: Causal and Conceptual Homogeneity in Qualitative Research authored by Mahoney, James.

Case study researchers will often express real concern for this aspect of homogene-

ity. A scholar with expertise in a particular region may look at various instances of 5
and worry whether they are really the same. Case expertise may lead the analyst to see
concept heterogeneity where a large-N dataset assumes homogeneity.

For example, Bowman, Lehoucq, and Mahoney (2005) argue that the use of inap-

propriate data sources causes the leading quantitative datasets measuring democracy
(i.e., Gasiorowski 1996; Marshall and Jaggers 2004; Vanhanen 2004) to seriously mis-
code the Central American countries. Because of poor underlying data, two cases may
have the same score for democracy across measures but actually have quite diﬀerent
levels of democracy. Indeed, these authors’ own assessment of the level of democracy
in Central America (the BLM index) produces results that are only weakly correlated
with existing indices (and for certain periods there is no correlation at all).

From the perspective of the BLM index, then, the scores in the other measures

often suﬀer from conceptual heterogeneity. Because they lack reliable data, the same
score across country-years reﬂects a very diﬀerent underlying reality about level of
democracy. For instance, Gasiorowski (1996) codes Nicaragua 1981 and Nicaragua
1985 both as zero (authoritarian). However, Bowman, Lehoucq, and Mahoney argue
that these zeros are not equivalent; the zero for 1981 is appropriate, but the zero for
1985 underestimates the impact of the elections of 1984 on the quality of democracy
in this country.

Conceptual homogeneity concerns also arise through what we call “substitutabil-

ity” (Goertz and Mahoney 2005) or what might be called functional equivalence (Prze-
worski and Tuene 1970; van Deth 1998). The Polity IV concept of democracy is an
aggregation of ﬁve component parts. Hence, there are a variety of ways to score a 5 on
the polity scale. Concept homogeneity assumes that in terms of causal eﬀect all these
conﬁgurations are homogeneous. The key assumption is that some diﬀerences within
democracies have no causal eﬀect and hence can be treated as equivalent. Przeworski
et al. (2000) provide an illustration of this homogeneity assumption. Their concept of
democracy has four component parts. Their coding rules state that if a country has
a zero value (dichotomously) on any one of the four components, then the country
is coded as a nondemocracy. Democracy can be achieved in only one way (i.e., a one
on all four components, whereas nondemocracy can occur in 15 diﬀerent ways (i.e.,
2

−

1 = 15).

The question is whether these 15 diﬀerent ways of becoming a nondemocracy have

the same causal consequences when introduced into analysis (or, alternatively, are
produced by the same causal factors). For example, when assessing the consequences
of nondemocracy on fertility rates, as Przeworski et al. (2000) do, can we assume that
a country that has zero value on only one of the components performs in the same
way as a country that has a zero value on all four components?

This kind of homogeneity issue is especially important when scholars are work-

ing with dichotomous categories, which is common in case study research. Quite