panel regression / fixed effects / time fixed effects as well

I have panel data including 48 US states from 1900-1920. I am estimating a fixed effects model with heteroscedasticity and autocorrelation:
xtreg y x, fe vce (cluster state)
A time series plot of y aggregating across states shows a positive linear trend. The testparm command indicates there are time-fixed effects. My question is do I then have to estimate the model using 20 dummy variables as:
xtreg y x i.year, fe vce (cluster state)
Or is it possible to include the single continuous variable, year, and use:
xtreg y x c.year, fe vce (cluster year)
Thanks, Robert Biggert