Why the author is quite loose when controlling in control variables for exchange rate regression?

Question

I read a paper from (Han 2020) and from his equation (4)

Δ s i , t + h = α i + β 1 x j , t + β 2 Δ π i , t + h + ε i , t + h , t = 1 , 2 , ⋯ , T − h

where xj,t is the change in Baltic Dry Index, Δπi,t is the inflation differential, and Δsi,t+h is the change in exchange rate.

However, I saw that the exchange rate also be affected by interest rates, confidence, the current account on balance of payments, economic growth and relative inflation rates. So why the paper here did not account for these control variables?

Answer 1

There is an important conceptual econometric distinction between identifying causal effects and predicting a variable. When you try to identify causal effects, it's important to control for confounders like the ones you mentioned. Otherwise, you may erroneously attribute the effect of the omitted variables to the Baltic Dry Index at hand. But the referenced paper is concerned with "predictive ability". Here, you do not care why the Baltic Dry Index helps predicting the exchange rate. It may well be that it proxies for e.g. confidence. But if your goal is only to explain the exchange rate given the Baltic Dry Index, you may not even need to observe confidence, because the Baltic Dry Index may be sufficient to capture the effect. Of course, you may be interested in the added forecasting ability beyond the effect of confidence. For that partial effect, you would need the other variables.