For quite some time I was searching for older population numbers than those of the 1990 census I could find within the reports of the 2000 census - and even those data were incomplete as for some of the provinces created in 1993 there were no data for 1990 mentioned. At least at province level, there are the data collected by Gwillim Law on his statoids website, but for 1960 and 1970 there are only rounded numbers. There are a few other online resources, like this 1970 report, and some pages on the Chulalongkorn university webserver from a study about the number of elderly people. While I was able to collect the census data at province level with those resources, there seemed to be no resources online which go down to district level, the level published in the census reports by the National Statistical Office (NSO). I have the reports for 2000 and 2010, and one from 1960, but nothing else.
National Library which has all of these books - but unfortunately they are in a major reconstruction and all of the old books are stored away and not accessible. Final attempt was to ask my sister in law who has access to a university library, but instead of getting me the books she came up with something much better.
The NSO has a online database of scanned publication, and most of the documents are exactly those census reports which I was searching for. As a slight drawback that website has the individual pages as PNG files, and not whole publications as one PDF, but that's not so difficult with some scripting and free tools. In the past week I have already filled in the 1990 data where I either had absolutely none - the districts of Nong Bua Lam Phu and Sa Kaeo - or where the data I had did not sum up to the values of the provinces. But the amount of data - not just the total population, but also those in private and collective household, the agricultural population, and that separately for municipal, sanitary district area and the other areas - makes it a quite tedious and time-consuming work, even with an Excel sheet to help spotting typing mistakes and doing the conversion to XML.
Instead of continuing with 1990, I instead started 1960, as that will be much faster to complete. Not only there were fewer districts in 1960 than 1990, also the reports give much less details - just the total and the municipal population, and the agricultural and non-agricultural households. The south is already done in the corresponding XML, and the other regions should be done within the next month. Whereas the amount of data for 1970 is similar with 1960, the reports for 1980 and 1990 will keep me busy at least a year unless I can find some help...