Between-Groups Estimator

The between-groups estimator uses group averages in panel data to estimate how outcomes differ across entities rather than within each entity over time.

The between-groups estimator is a panel-data estimator that runs a regression on group means, so it explains differences across entities rather than changes within each entity over time.

$$$$

How it works

Suppose panel data are indexed by entity (i) and time (t). The between estimator first averages the data over time for each entity:

$$ \bar{y}_i = \alpha + \beta \bar{x}_i + \bar{u}_i $$

It then applies ordinary least squares to those averaged observations.

This means the estimator uses between-unit variation only. If a country has a higher average tax rate and lower average growth than another country, that cross-country difference contributes to the estimate. Year-to-year variation inside each country does not.

Why it matters

The between estimator is useful when the researcher wants to study long-run differences across groups, but it cannot control away time-invariant omitted characteristics the way a fixed-effects estimator can.

So the main trade-off is:

  • easier interpretation of cross-group differences,
  • but more vulnerability to omitted-variable bias from unobserved group traits.

Knowledge Check

### What source of variation does the between-groups estimator use? - [x] Differences in time averages across entities - [ ] Only year-to-year variation within each entity - [ ] Only random sampling variation with no panel structure - [ ] Only variation in the error term > **Explanation:** The estimator collapses each entity to its average values and estimates the relationship across those averages. ### What is a main limitation of the between estimator? - [x] Unobserved time-invariant differences across groups can still bias the estimate - [ ] It cannot be used with panel data - [ ] It ignores cross-sectional variation completely - [ ] It always has lower variance than fixed effects > **Explanation:** Because it compares groups rather than differencing them out, omitted group traits can remain in the error term. ### Which estimator is the natural contrast to the between estimator in panel-data analysis? - [x] Within-groups estimator - [ ] Treasury-bill estimator - [ ] Beveridge estimator - [ ] Bid-ask estimator > **Explanation:** The within estimator focuses on changes inside each entity over time, which is the opposite source of variation.