Hi,

I am trying to explain the formation of shipping freight rates through the use of several different variables in a random effects panel data model. The explanatory variables consist of ship specific variables (such as size of the ship), contract specific variables (such as price and date of the contract) and macro variables (such as the general market shipping rate and the crude oil price). The macro variables only vary over time, not ship, while the other two groups of variables vary across ships only or both time and ship.

I am considering adding time dummies to account for time passing in the model. I have tried adding daily time dummies, but many of these are omitted due to perfect collinearity with the market rate. I then tried with monthly dummies, and none were omitted. I believe this is because the monthly dummies are changing at a different "pace" than the market rate (which changes on a daily basis). When adding these monthly dummies the model increases in R2 and my explanatory variables obtain higher significance and coefficients. However, I am worried that these results are artificially good due to including both time dummies and the macro variables (especially the market rate). Could this be the case or can I actually trust these improved results?

I appreciate any help on this matter.