Investigating regional effects

[Mark McCarthy]

Hi,

I am currently investigating the effects of labour demand on house prices at a local authority level. I have my data in yearly (%) changes for the period 2016-2021, for 320 local authorities in the UK.

I have allocated each local authority an ID - which range from 1-320.

I have been using time and region fixed effects using the functions (i.year i.ID), but have since been suggested to use i.year, fe cluster(ID). The results are similar for both, but using (i.year i.ID) gives me a higher R squared value, I am wondering if that means I should use this command over the other?

Also, I have then categorised each local authority into 1 of 9 wider regions (south east, London, north east etc..), so each ID belongs to a group 'region' in which they are labelled with a number (1-9) depending on the region they belong to.

I am looking to investigate whether the effects of labour demand vary across these 9 regions, and was hoping someone could advise me on how to do so? I have tried using c.labour_demand##region, which gives me some results, but I am unsure on whether that is telling me what I am trying to investigate?

thank you very much, any help would be really appreciated.

[George Ford]

could use
reghdfe y x1 x2 , absorb(year ID) cluster(ID)

The R-squared is not a basis to select an estimation method. The results should be identical except for a small differences in the standard errors due to "degrees of freedom" difference. xtreg demeans everything before estimation, which avoids reducing df with a bunch of dummies.

I'd think the interaction term is giving you want you want. The interaction portion of it tells you the difference in the slope of one region versus the base region and the t-stat is a test statistic on the null hypothesis = 0. So if the coef on labour_demand was 1.0 and one of the interaction coefficients is -0.2, then in that region labour demand has a smaller effect on Y (20%) than in the base region.

[Mark McCarthy]

Thank you for the response George!

I have seen the reghdfe mentioned before and have tried it, it seems to be unrecognised in my STATA, is there a download I must complete before using it?

[Mark McCarthy]

Also, in regards to the second part of the question: is the base year the year that does not appear with the rest of the regions (seems to be omitted) - if so, is the coefficient displayed on labour demand above referring to this base region?

[Mark McCarthy]

sorry I have found the installation and reghdfe is now working for me, thanks!