Short Run vs Long Run Effect in Dynamic Panel Regressions This video differentiates between short run and long run effects of an independent variable in dynamic panel regression (from 19:25 to 20:50). Firstly, I would like to know when and why do we differentiate between these effects. Secondly, which effect is usually sensible to be reported in research papers (I believe majority of research papers focus on the short run effects because they interpret the coefficients directly without any transformation).
 A: We care about dynamic models (time series models, cointegrating regressions, ADL models, etc.) because we want to model the dependent variable's memory of itself. Put another way, we want our model to include a function of past values of the dependent variable as a predictor. When a variable has a non-zero memory, a few things can happen:


*

*The memory is short, and perturbations to the dependent variable rapidly return to a stable equilibrium.

*The memory is long, and perturbations to the dependent variable take a long or very long time to return to a stable equilibrium.

*The memory is infinite, but not an explosive positive feedback type (i.e. i.e. it has unit root), and perturbations to the dependent variable never cease influencing future values.

*The memory is infinite and of a positive feedback variety where older perturbations have the strongest influence on future values of the dependent variable.
There is no general answer to your question about which to report, as the substantive nature of the inquiry, study design, analytic model, etc. will bear strongly, as will the nature of the dependent variable's memory process. However, here's why you might care about different kinds of effects on outcomes with memory processes:
Long-run effects: One may care about what is happening to a stable equilibrium (mean or mean trend). For example, is the long-run effect of a policy or signal to shift the equilibrium level or trend (i.e. the level over time) up or down?
Short-run effects: First of all, there are two kinds of short-run effects: instantaneous, and lagged. About them:
Instantaneous short-run effects: One may care about the effect of change in a predictor at a time period on the dependent variable at that time period. For example, is the instantaneous effect of first implementing a policy or the modulation of a signal to shift the dependent variable at that period up or down?
Lagged short-run effects: One may care about the effect of the level in a predictor at a previous (lagged) time period on the dependent variable at the current time period. For example, is the lagged short-run effect of the level of a  policy (e.g., spending per capita) or the level of a signal at a previous period to shift the dependent  variable at the current period up or down?
