Recoding income variables in trends analysis

[Esther Lamidi]

I will greatly appreciate your advice.

I will like to control for family income in my analysis of trends in self-rated health between year 2000 and 2018. The NHIS data from IPUMS that I'm using has five imputed family income variables. The imputed variables are measured in intervals, rather than point estimates. For years 2000-2007, some of the intervals are $5,000 apart while others are $10,000 apart. The last category for years 2000-2007 is $75,000 and above (please see below). For the latter years, the intervals are mostly $5,000 apart and the categories extend beyond $75,000 and above, to around $115,000 and above. This means that whereas the income variables for the first several years have 11 categories, the latter years have up to 24 categories. My question is, which of the two approaches below would you recommend that I use?

1. Convert the midpoint of each interval of the imputed income variables to 2018 dollars and treat the new variable as a continuous variable (of course analyzing all five datasets with the imputed five variables using mi estimate in stata)
2. Measure income in few categories (e.g. <$25,000, $25,000-$44,999, $45,000-64,999, $65,000-$74,999 and $75,000 and above) that will be consistent across years (also analyzing all five datasets with the imputed variables using mi estimate in stata)

The first approach will allow me to adjust for inflation, but I worry about the differences in the measurement of income across the survey years
I am not sure if it is okay to simply control for family income in categories in a trend analysis. Even if it is okay to control for family income in ordinal categories as in 2 above, I wonder if I will then be controlling for differences in income among respondents rather than changes in income over the years. Please help me. I am open to other suggestions.


Categories of imputed income variables for 2000-2007

imputed total
combined family
income (1997+
grouping)	Freq.	Percent	Cum.
$0-$4,999	495	5.11	5.11
$5,000-$9,999	1,013	10.46	15.58
$10,000-$14,999	943	9.74	25.32
$15,000-$19,999	901	9.31	34.62
$20,000-$24,999	814	8.41	43.03
$25,000-$34,999	1,331	13.75	56.78
$35,000-$44,999	1,070	11.05	67.83
$45,000-$54,999	870	8.99	76.81
$55,000-$64,999	660	6.82	83.63
$65,000-$74,999	489	5.05	88.68
$75,000 and over	1,096	11.32	100.00

Categories of imputed income variables for 2007-2018

imputed total
combined family
income (1997+
grouping)	Freq.	Percent	Cum.
$0-$4,999	184	1.92	1.92
$5,000-$9,999	273	2.86	4.78
$10,000-$14,999	450	4.71	9.49
$15,000-$19,999	431	4.51	14.00
$20,000-$24,999	553	5.78	19.78
$25,000-$29,999	560	5.86	25.64
$30,000-$34,999	638	6.67	32.31
$35,000-$39,999	491	5.14	37.45
$40,000-$44,999	567	5.93	43.38
$45,000-$49,999	477	4.99	48.37
$50,000-$54,999	459	4.80	53.17
$55,000-$59,999	362	3.79	56.96
$60,000-$64,999	445	4.65	61.61
$65,000-$69,999	206	2.15	63.77
$70,000-$74,999	337	3.53	67.29
$75,000-$79,999	330	3.45	70.74
$80,000-$84,999	304	3.18	73.92
$85,000-$89,999	147	1.54	75.46
$90,000-$94,999	191	2.00	77.46
$95,000-$99,999	131	1.37	78.83
$100,000-$104,999	314	3.28	82.11
$105,000-$109,999	91	0.95	83.06
$110,000-$114,999	106	1.11	84.17
$115,000 and over	1,513	15.83	100.00