Good Afternoon,

I am currently undertaking a piece of research in which I require a fixed-effects model. I have panel data in the form of 6 years & 48 Local Authorities giving 288 obs.

My research is looking at estimating the impact of immigration on house prices.

xtset LA_id Year
panel variable: LA_id (strongly balanced)
time variable: Year, 2013 to 2018
delta: 1 year

The Issue:
When using a FEIV Model, when i dont include year dummies via i.year i get results with a significant coefficient on the IV:

xtivreg Change_lnHPI Change_in_Population_by_Initial lag_Housing_Population UN100 CC100 (ActualStockChange_lagPopStock = Flow_Population), fe vce(cluster GOR_id)

Fixed-effects (within) IV regression Number of obs = 288
Group variable: LA_id Number of groups = 48
Coefficient on IV = 3.99 significant at a 5% level

However, when i introduce i.Year into the model to capture year effects (2013 as a base year) the coefficient on my IV is extremely insignificant...

I am wondering if anyone can help to offer a justification in this situation as to why I should exclude year effects from the model... It's hard to argue that year dummies aren't necessary when considering something like house prices which are so clearly time-dependent.

Note: When using an IV2SLS model, coefficients are perfectly fine with the inclusion of year dummies, so this makes this result even more baffling.

Thank you.