I am modeling how entities dynamically reallocate emphasis across three related topics (A, B, C) over time. The topics represent shares of an overall activity, so increasing emphasis in one may reduce emphasis in others.
I have a panel dataset tracking how entities allocate emphasis across three related topics (A, B, and C) over time (around 8 years of data). For each entity–year, I observe the relative emphasis placed on each topic. I also observe the timing of exogenous, entity-specific shocks (e.g., some event related to a certain topic affecting that entity).
I want to study how shocks change the way emphasis is distributed across the three topics over time. In particular, I want to see how a shock affects the topic it directly hits and whether it also changes emphasis in the other topics. I also expect there may be trade-offs between topics (for example, putting more emphasis on one may reduce emphasis on the others). What would be a good way to model and test these dynamic effects in panel data?
I was thinking about using a dynamic panel model that includes lagged values of the topics and the shock variable, possibly written in an error-correction form so I can distinguish short-run from long-run changes in allocation. I would estimate one equation for each topic with entity fixed effects. Would this be appropriate, or what would be a sensible approach?
I have a panel dataset tracking how entities allocate emphasis across three related topics (A, B, and C) over time (around 8 years of data). For each entity–year, I observe the relative emphasis placed on each topic. I also observe the timing of exogenous, entity-specific shocks (e.g., some event related to a certain topic affecting that entity).
I want to study how shocks change the way emphasis is distributed across the three topics over time. In particular, I want to see how a shock affects the topic it directly hits and whether it also changes emphasis in the other topics. I also expect there may be trade-offs between topics (for example, putting more emphasis on one may reduce emphasis on the others). What would be a good way to model and test these dynamic effects in panel data?
I was thinking about using a dynamic panel model that includes lagged values of the topics and the shock variable, possibly written in an error-correction form so I can distinguish short-run from long-run changes in allocation. I would estimate one equation for each topic with entity fixed effects. Would this be appropriate, or what would be a sensible approach?
Comment