Working Paper

Should We Care (More) About Data Aggregation? Evidence from Democracy Indices.

Klaus Gründler, Tommy Krieger
CESifo, Munich, 2019

CESifo Working Paper No. 7480

Aggregation tools transform multidimensional data into indices. To investigate how the design of an aggregation process influences regression results, we build democracy indices that differ regarding their scale and aggregation function. Our results show that the shape of the aggregation function significantly affects OLS and 2SLS estimates since different shapes produce systematically different index values for regimes at the lower and upper end of the democracy spectrum. We also find that dichotomous indices produce significantly smaller OLS estimates than continuous indices because of greater measurement uncertainty. Whether continuous and dichotomous indices cause different 2SLS estimates depends on their design.

CESifo Category
Fiscal Policy, Macroeconomics and Growth
Empirical and Theoretical Methods
Keywords: aggregation, data transformation, democracy, economic development, indices, machine learning, measurement of democracy, political transitions, scaling
JEL Classification: C260, C430, O100, O430, P160, P480