Why won't 'twoways' in plm produce fixed effects for group as well as time?

Question

When I do the two ways model in the plm package, my understanding is that it should have fixed effects for group and time, but when I manually look at the fixed effects it only produces fixed effects for the group variable. For example, using the canned data in the plm package:

> data("Produc", package = "plm")
> zz <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
            data = Produc, index = c("state","year"),model='pooling')
> qq <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
            data = Produc, index = c("state","year"),model='within',effects='twoways')
> length(fixef(zz))
Error in fixef.plm(zz) : fixef is relevant only for within models
> length(fixef(qq))
[1] 48
> length(unique(Produc$state))+length(unique(Produc$year))
[1] 65

My expectation is that the last two lines should be equal, i.e. that there should be year and state fixed effects. Why are they different?

Helix123 Helix123 · Accepted Answer · 2019-08-16T21:25:16

This is explained in the help for fixef, use ?fixef to view it:

For a two-ways model, fixef's default behaviour is to output the group ("individual") fixed effects. If you want the time fixed effects, use fixef(your_model, effect = "time").

Picking up your example, take heed you do not estimate a two-way fixed effects model as the argument is called effect rather than effects (with a s). So use:

qq <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
          data = Produc, index = c("state","year"), model ="within", effect = "twoways")
length(fixef(qq)) # 48
length(fixef(qq, effect = "time")) # 17

You can check with summary(qq) which model you estimated as the first line of its output will tell you the model: "Twoways effects Random Effect Model" in this case.

Why won't 'twoways' in plm produce fixed effects for group as well as time?

1 Answers