The Study Design. To estimate an elasticity, researchers ideally would observe an exogenous change in either price or income, with no change in any other factor that would affect demand. However, researchers generally have lacked this ideal situation and have taken various alternative approaches to estimating elasticities. A study that uses a random-assignment experimental design (the most sound but costly way to observe response to change) is likely to produce a different estimate than a study that uses a quasi-experimental or observational design.
The Data Source and Study Period. Elasticity estimations may be derived from
various types of data—primary sectional or longitudinal, local
secondary, survey or administrative, cross-
limitations. They may define key variables types and magnitudes of measurement error.
differently and also introduce different Largely determined by available data,
the study period may also affect estimates of elasticity. In general, estimates more recent data are more likely to be more relevant for policy development.
The Empirical Method. Studies that define demand and price in the same way, calculate elasticity in the same way, and observe the same time period with the same source of data still can produce very different estimates if they use different empirical methods. Researchers have developed econometric models of increasing complexity to deal with problems such as selection bias, a problem commonly associated with observational data.
Considering the many factors that affect the estimation of elasticity, no single estimate of
price or income elasticity is likely to apply in all circumstances and to all populations. Instead,
differences in how estimates are calculated typically limit how they should be applied in
modeling policy proposals and in drawing conclusions. In this report, we attempt to define the
range of elasticity estimates offered in the research literature, with respect to health insurance
and health care services generally, and for specific products and subpopulations.