Can you test the parallel trends assumption?

Can you test the parallel trends assumption?

Parallel Trend Assumption Although there is no statistical test for this assumption, visual inspection is useful when you have observations over many time points.

What are parallel trends?

The parallel trends assumption states that, although treatment and comparison groups may have different levels of the outcome prior to the start of treatment, their trends in pre-treatment outcomes should be the same.

Are parallel trends testable?

Second, parallel pre-intervention trends is not an assumption at all! It is a testable empirical fact about the pre-intervention outcomes, involving no counterfactuals.

What are common trends assumptions?

Common Trend Assumption: AA represents the counterfactual trend, or parallel or common trend for the treatment group that would occur in absence of treatment.

When should we use Difference in Difference?

The difference-in-difference method captures the significant differences in outcomes across the treatment and control groups, which occur between pre-treatment and post-treatment periods. In the simplest quasi-experiment, an outcome variable is observed for one group before and after it is exposed to a treatment.

What is an identifying assumption?

Identifying assumption: assumptions made about the DGP that allows you to draw causal inference. In other words, the ‘identification assumption’ you make for estimate the causal effect of smoking on cancer rates, i.e. that smokers & non-smokers only differ in terms of their smoking behavior, is likely not to hold here.

How to report treatment effect of parallel trends?

The idea here is to first report treatment effect estimates from a DiD model with a more complex trend difference than you think to be the case – so if you think there are parallel trends, first estimate a model that allows for the two groups to have linear trends with different slopes by including a linear trend difference.

Which is the optimal flci for parallel trends?

For these types of restrictions, results fromArmstrong and Kolesar(2018, 2020) imply that the optimal FLCI has near-optimal expected length when in fact parallel trends holds. Unfortunately, however, we show that FLCIs can have poor properties for broader classes of restrictions, such as those that incorporate sign or shape restrictions.

What makes a large difference in parallel trends?

A key issue here is determining what constitutes a “large’ difference – which takes you back to the type of graph shown above, which examines robustness to different degrees of deviation from parallel trends. The authors’ R code is under construction, and will be available here.

When to use sensitivity analysis for parallel trends?

Importantly, these methods should be used when there is reason to be skeptical of parallel trends ex ante, regardless of the outcome of a test of whether parallel trends hold pre-intervention. This type of sensitivity analysis will allow one to get bounds on likely treatment effects.