How to summarize the income earned out of several income sources

[Linda WUR]

Dear all,

I have an urgent question with regards to summarizing the income earned out of different income sources. To our treatment and control group we have asked what they earn from several income generating activities. I want to be able to compare the differences in income between the treatment and control group. The challenge is that the income data is labeled, as the repsondents were given a range in which they could answer. These ranges made it easier for the targeted group to answer, but more difficult for me now to analyse the data.

The ranges given per income generating activity were:
0 - 125.000 (label = 1)
125.005 - 250.000 (label = 2)
250.005 - 500.000 (label = 3)
500.005 - 1.000.000 (label = 4)
> 1.000.000 | (label = 5)
Don't know | (label = 999)


I tried to do the following

egen AIGA_averagemonth = rowtotal( trade1_ic_month trade2_ic_month trade3_ic_month trade4_ic_month trade5_ic_month trade6_ic_month trade7_ic_month trade8_ic_month trade9_ic_month trade10_ic_month agri_ic_month ls_ic_month proces_ic_month cook_ic_month )

But this did not lead to results I can interprete

[Clyde Schechter]

With the data aggregated into categories like this you cannot calculate an average income, nor a median or other such statistic. While you will sometimes see people use the middle of the range in each category and impute that value to everyone in the category, that is really not giving a valid impression of the income distribution. For the purposes of comparing income distributions by category between groups, you can use the categorized variable as it is and just do a cross-tabulation along with, say, a chi square statistic. -tab income_variable group, chi2- will do that.

You can't really get a handle on total income from this kind of data either. Even if you know that person A had more category 1 income and less category 2 income than person B, either one of them could have had the larger total income. Only when one person's income exceeds another person's income in every category can you draw a conclusion about the relative order of their total incomes. But that will still leave you uncertain in a large number, perhaps even a majority, of cases.