Projection/ Forecasting of Population census data in stata

[Radhika Channanamchery]

Hi,

I want to project population census data for a particular year. I have the census data for the periods 2001 and 2011. I need population data for the years 2005 and 2015. How do I project or forecast this in Stata?

Any help would be greatly appreciated.
Thank you
Radhika

[Hemanshu Kumar]

The command ipolate allows you to easily interpolate and extrapolate data in a linear fashion. If you want to do geometric interpolation, you can do that by first taking logs of the data and then interpolating linearly using ipolate.

Forecasting future values of the population is generally a tricky thing. For India, you can take a look at this document to see how the Registrar General's office itself does it; you may even want to use their numbers directly rather than coming up with your own.

[Radhika Channanamchery]

The command ipolate allows you to easily interpolate and extrapolate data in a linear fashion. If you want to do geometric interpolation, you can do that by first taking logs of the data and then interpolating linearly using ipolate.

Forecasting future values of the population is generally a tricky thing. For India, you can take a look at this document to see how the Registrar General's office itself does it; you may even want to use their numbers directly rather than coming up with your own.

Dear Hemanshu Kumar,

Thank you very much for your kind reply.

Thank you for sharing the document. I am working with the following district level data.

District	urban_pop_2001	urban pop_2011
1	478520	597846
2	407396	536503
3	5759987	8749944
4	1012691	1211195
5	706850	920239
6	344875	425952
7	396089	501978
8	148090	174974
9	222724	239508
10	274238	329533
11	729302	996086
12	542998	629010
13	881917	1049539
14	342183	379309
15	852621	1056228
16	304673	376763
17	299020	355501
18	75382	80988
19	625523	761192
20	198292	233704
21	282715	308362
22	982128	1245413
23	420837	490348
24	571010	623727
25	507202	599078
26	206353	334061
27	387913	418981

How will I give interpolate command here? What is the right specification? I want to know the projected values for the years 2005 and 2015.

I would be very grateful if you could help me with this.

Thank you.

Radhika

[Hemanshu Kumar]

A simple geometric interpolation and extrapolation can be done as follows:

clear
input byte district long(urban_pop_2001 urban_pop_2011)
1    478520    597846
2    407396    536503
3    5759987    8749944
4    1012691    1211195
5    706850    920239
6    344875    425952
7    396089    501978
8    148090    174974
9    222724    239508
10    274238    329533
11    729302    996086
12    542998    629010
13    881917    1049539
14    342183    379309
15    852621    1056228
16    304673    376763
17    299020    355501
18    75382    80988
19    625523    761192
20    198292    233704
21    282715    308362
22    982128    1245413
23    420837    490348
24    571010    623727
25    507202    599078
26    206353    334061
27    387913    418981
end

gen long urban_pop_2005 = .
gen long urban_pop_2015 = .

reshape long urban_pop_ , i(district) j(year)

gen ln_urban_pop_ = ln(urban_pop_)

ipolate ln_urban_pop_ year, gen(i_ln_pop) epolate by(district)
replace urban_pop_ = exp(i_ln_pop) if inlist(year, 2005, 2015)

drop i_ln_pop ln_urban_pop_

reshape wide

However, I want to underline that I do not really recommend this method for extrapolation to 2015: those projections should not be based on just 2001 and 2011 overall population numbers.

[Radhika Channanamchery]

A simple geometric interpolation and extrapolation can be done as follows:

clear
input byte district long(urban_pop_2001 urban_pop_2011)
1 478520 597846
2 407396 536503
3 5759987 8749944
4 1012691 1211195
5 706850 920239
6 344875 425952
7 396089 501978
8 148090 174974
9 222724 239508
10 274238 329533
11 729302 996086
12 542998 629010
13 881917 1049539
14 342183 379309
15 852621 1056228
16 304673 376763
17 299020 355501
18 75382 80988
19 625523 761192
20 198292 233704
21 282715 308362
22 982128 1245413
23 420837 490348
24 571010 623727
25 507202 599078
26 206353 334061
27 387913 418981
end

gen long urban_pop_2005 = .
gen long urban_pop_2015 = .

reshape long urban_pop_ , i(district) j(year)

gen ln_urban_pop_ = ln(urban_pop_)

ipolate ln_urban_pop_ year, gen(i_ln_pop) epolate by(district)
replace urban_pop_ = exp(i_ln_pop) if inlist(year, 2005, 2015)

drop i_ln_pop ln_urban_pop_

reshape wide

However, I want to underline that I do not really recommend this method for extrapolation to 2015: those projections should not be based on just 2001 and 2011 overall population numbers.

Thank you very much for the codes and for sharing your insights on this.