Extract year from quarterly dates

[Francois Durant]

Hi,

Seems like a very basic question but I can find the basic function that extract the year from quarterly dates

Here is what I get when I try:

gen year = year(qyear)

qyear    year
1986q2    1960
1986q3    1960
1986q4    1960
1987q1    1960
1987q2    1960
1987q3    1960
1987q4    1960
1988q1    1960
1988q2    1960
1988q3    1960
1988q4    1960
1989q1    1960
1989q2    1960
1989q3    1960
1989q4    1960
1990q1    1960
1990q2    1960
1990q3    1960
1990q4    1960
1991q1    1960
1991q2    1960
1991q3    1960
1991q4    1960
1992q1    1960
1992q2    1960
1992q3    1960

and I understand why...

I would have liked to use one of the commands "qofd" but I could not find the of for quarterly date ("invalid syntaxt"):

gen year = yofq(qyear)

Is there any way beside converting to string and extract the date from string? Thanks!

[Jorrit Gosens]

Please use dataex. That will allow people to see the current format of your qyear variable, which matters for what the correct answer is. It also allows people to copy paste (the above cant).
See also the explanation in the FAQ (http://www.statalist.org/forums/help#stata).

  ssc install dataex
  dataex in 1/20

[Nick Cox]

Please use dataex (SSC) to give data examples.

This is all documented at help dates or http://www.stata.com/help.cgi?dates: you just to have read carefully!

What are displayed in your example as e.g. 1986q2 are quarterly dates. At first sight they might be strings too, but your syntax would not then work.

I will use display with examples to which anyone should be able to work out an answer independently.

With the exception of calendar years, Stata uses 0 as origin for the first possible date in 1960.

So 0 is

* when interpreted as a daily date, 1 January 1960 

. di %td 0
01jan1960

* when interpreted as a a quarterly date, the first quarter of 1960 

. di %tq 0
1960q1

With 0 as origin a little mental arithmetic shows that quarterly dates in the mid-1980s are numerically about 25 x 4 quarters from the origin. A precise example is that the second quarter of 1986 is

. di yq(1986, 2)
105

. di %tq  yq(1986, 2)
1986q2

year() is a function to extract the calendar year from daily dates.

Hence it should not be a surprise that 1960 is returned for 105 as argument, as 105 is a daily date within 1960.

. di year(105)
1960

. di %td 105
15apr1960

This is documented: http://www.stata.com/help.cgi?year()

Francois understands this, but it's also documented (by absence) that Stata doesn't provide all possible date-to-date conversion functions. Perhaps it should and then we wouldn't have to write answers like this.

Hence you need to chain functions together: extract a daily date (strictly the beginning of each quarter) for each quarter, then extract a year:

. di yofd(dofq(105))
1986

[Francois Durant]

Nick and Jorrit,

Thanks for both answers. Nick's answer only work if you keep the same data sample since you input "105" in the answer. How would you code that in a way that I output the proper answer if I keep on changing the time series? For instance, I may run the same Stata code on a sample that start in 1979 without having the copy/paste the first date? Please let me know. Thanks!

[Francois Durant]

Sorry I responded too quickly. I can figure out the first year the following manner:

di yofd(dofq(qyear)) However, would I use this command to generate "year" for the whole sample? Thanks

[Jorrit Gosens]

. list

     +-------+
     |  var1 |
     |-------|
  1. |   105 |
  2. |  200  |
  3. | 12856 |
     +-------+

. gen year = yofd(dofq(var1))

. list

     +--------------+
     |  var1   year |
     |--------------|
  1. |   105   1986 |
  2. |   200   2010 |
  3. | 12856   5174 |
     +--------------+

that might not work for your data, depending on the current format. So again, please use dataex doing

  ssc install dataex
  dataex in 1/20

[Nick Cox]

I didn't think I needed to explain how to use generate !!!

[Francois Durant]

Apologies. I'm not familiar with the command "di". I should have tried gen...!

[Nick Cox]

I explained that I was using display:

I will use display with examples to which anyone should be able to work out an answer independently.

di is an abbreviation of display

[Obaid Ur Rehman]

Reply to #8
use the following command

gen year = yofd(dofq( qyear ))