I'm a statistician/machine learning scientist more familiar with molecular bio than economics. Trying to find out if an issue I perceive in bio modeling also occurs in econometric modeling.
A common task in computational biology is to build a model of some biochemical process from data. These biochemical processes are always dynamic, i.e. something you might model with an ODE or Kalman filter. However, there is a bias towards modeling such a process from equilibrium data. This means, you wait until the process has reached some equilibrium (homeostasis), and you take measurements across subjects/replicates. This is because for many reasons it is much easier and cheaper to acquire equilibrium data relative to time series data. Training a model only on equilibrium data is a problem of course if you want the model to be predictive when the system is not at equilibrium.
I wonder if the same bias happens with econometric models? Clearly, in terms of statistical methodology, cross-sectional data is typically cheaper and easier to acquire than time-series data. And I know economists are concerned with modeling equilibria. I'm hoping there is a place someone can point me to that talks about this issue in the econometric context, perhaps with an example?