In a comment at Overcoming Bias, Michael Vassar responded to the claim that in our society on average the wealthy are least promiscuous: “If you have spent time with middle class and wealthy people, even a little time, that claim is obviously false. The wealthy are MUCH more promiscuous in my observation.”
I decided to check the GSS.

Frequency Distribution
Cells contain:
-Column percent
-Weighted N
RINCOME
1
LT $6000
2
$6000 – 14999
3
$15000 – 24999
4
$25000 OR MORE
ROW
TOTAL
PARTNERS 0: NO PARTNERS 16.8
383.4
13.9
439.1
12.7
444.0
9.7
737.1
12.1
2,003.6
1: 1 PARTNER 65.2
1,482.9
69.5
2,189.1
73.2
2,565.4
79.1
6,000.6
74.1
12,237.9
2: 2 PARTNERS 8.6
195.8
7.5
236.5
7.3
255.9
5.3
402.5
6.6
1,090.7
3: 3 PARTNERS 3.5
80.6
4.2
132.9
3.2
112.8
2.5
188.6
3.1
514.9
4: 4 PARTNERS 2.4
55.4
2.0
61.9
1.7
58.2
1.5
112.3
1.7
287.8
5: 5-10 PARTNERS 2.1
48.6
2.2
70.1
1.3
46.7
1.4
109.6
1.7
275.0
6: 11-20 PARTNERS 1.0
23.1
.4
11.5
.3
10.8
.3
22.1
.4
67.6
7: 21-100 PARTNERS .2
4.0
.2
7.5
.2
8.0
.1
10.0
.2
29.5
8: MORE THAN 100 PARTNERS .1
1.6
.1
2.4
.0
1.5
.0
.5
.0
6.0
COL TOTAL 100.0
2,275.5
100.0
3,151.0
100.0
3,503.4
100.0
7,583.2
100.0
16,513.2
Means 1.21 1.21 1.15 1.13 1.16
Std Devs 1.12 1.04 .92 .85 .95
Unweighted N 2,062 2,987 3,471 7,550 16,070
Frequency Distribution
Cells contain:
-Column percent
-Weighted N
RINCOME
1
LT $3000
2
$3000 – 5999
3
$6000 – 9999
4
$10000 – 14999
5
$15000 – 19999
6
$20000 – 24999
7
$25000 OR MORE
ROW
TOTAL
PARTNERS 0: NO PARTNERS 18.2
203.5
15.5
179.9
15.6
199.6
12.8
239.4
13.8
230.5
11.7
213.5
9.7
737.1
12.1
2,003.6
1: 1 PARTNER 63.0
704.0
67.3
778.9
69.0
884.9
69.8
1,304.2
71.0
1,186.2
75.3
1,379.1
79.1
6,000.6
74.1
12,237.9
2: 2 PARTNERS 8.2
91.9
9.0
103.9
8.0
103.0
7.2
133.6
8.1
135.6
6.6
120.3
5.3
402.5
6.6
1,090.7
3: 3 PARTNERS 3.7
41.8
3.4
38.8
4.2
54.1
4.2
78.7
3.2
53.7
3.2
59.1
2.5
188.6
3.1
514.9
4: 4 PARTNERS 3.2
35.7
1.7
19.7
1.1
14.3
2.6
47.6
1.5
25.5
1.8
32.7
1.5
112.3
1.7
287.8
5: 5-10 PARTNERS 1.9
21.7
2.3
26.9
1.4
17.9
2.8
52.2
1.6
27.1
1.1
19.6
1.4
109.6
1.7
275.0
6: 11-20 PARTNERS 1.4
15.8
.6
7.3
.3
4.2
.4
7.3
.6
9.4
.1
1.4
.3
22.1
.4
67.6
7: 21-100 PARTNERS .2
1.9
.2
2.1
.4
5.3
.1
2.2
.2
2.6
.3
5.5
.1
10.0
.2
29.5
8: MORE THAN 100 PARTNERS .1
1.6
.0
.0
.0
.0
.1
2.4
.0
.4
.1
1.1
.0
.5
.0
6.0
COL TOTAL 100.0
1,118.0
100.0
1,157.5
100.0
1,283.4
100.0
1,867.7
100.0
1,671.1
100.0
1,832.3
100.0
7,583.2
100.0
16,513.2
Means 1.24 1.19 1.14 1.25 1.16 1.14 1.13 1.16
Std Devs 1.20 1.05 .97 1.09 .96 .88 .85 .95
Unweighted N 1,001 1,061 1,206 1,781 1,646 1,825 7,550 16,070

I shrunk the number of buckets because the GSS divides it too finely at the low end of the scale*. I wish I could have broken up the last bucket to reveal the really rich. At any rate, promiscuity seems to decrease with income, and Vassar didn’t distinguish between the rich and middle class anyway.
*I often forget how to rename, ranges, so here’s what I used: RINCOME(r: 1-5 “LT $6000”; 6-9 “$6000 – 14999”; 10-11 “$15000 – 24999”; 12 “$25000 OR MORE”). I don’t think I’ve mislabeled the dollar amounts, run it again without the aggregation into buckets if you’d like to check my boundaries. I also restricted the dependent variable with PARTNERS(0-8), since I don’t know how to treat answer 9.

Summary Statistics
Eta* = .04 Gamma = -.01 Rao-Scott-P: F(24,3168) = 8.37 (p= 0.00)
R = -.04 Tau-b = .00 Rao-Scott-LR: F(24,3168) = 8.10 (p= 0.00)
Somers’ d* = .00 Tau-c = .00 Chisq-P(24) = 256.99
Chisq-LR(24) = 248.81
*Row variable treated as the dependent variable.

On second thought, the problem with the income numbers is that inflation has reduced the values of the dollar categories they originally assigned. Breaking things up by YEAR I found that 1990 was the last year that the highest income category was comparable in size to my other buckets. So here are the results with the selection filter YEAR(1988-1990).

Frequency Distribution
Cells contain:
-Column percent
-Weighted N
RINCOME
1
LT $6000
2
$6000 – 14999
3
$15000 – 24999
4
$25000 OR MORE
ROW
TOTAL
PARTNERS 0: NO PARTNERS 17.5
90.9
11.1
73.4
12.5
81.2
7.4
58.7
11.6
304.3
1: 1 PARTNER 63.3
328.6
71.4
472.5
74.2
480.8
82.8
655.9
73.9
1,937.9
2: 2 PARTNERS 8.5
44.0
6.7
44.5
6.4
41.4
4.4
34.5
6.3
164.4
3: 3 PARTNERS 4.3
22.6
4.8
31.9
2.9
18.8
2.5
19.5
3.5
92.7
4: 4 PARTNERS 3.2
16.3
2.0
13.0
2.0
12.6
.8
6.3
1.8
48.3
5: 5-10 PARTNERS 1.2
6.4
3.2
21.5
1.7
11.1
2.1
16.8
2.1
55.7
6: 11-20 PARTNERS 1.2
6.4
.2
1.0
.2
1.6
.0
.0
.3
9.0
7: 21-100 PARTNERS .5
2.6
.6
3.7
.1
.5
.1
.5
.3
7.4
8: MORE THAN 100 PARTNERS .2
1.1
.0
.0
.0
.0
.0
.0
.0
1.1
COL TOTAL 100.0
518.9
100.0
661.5
100.0
648.0
100.0
792.2
100.0
2,620.7
Means 1.25 1.28 1.14 1.13 1.19
Std Devs 1.20 1.11 .91 .80 1.00
Unweighted N 470 633 639 786 2,528
Color coding: <-2.0 <-1.0 <0.0 >0.0 >1.0 >2.0 Z
N in each cell: Smaller than expected Larger than expected
Summary Statistics
Eta* = .07 Gamma = -.02 Rao-Scott-P: F(24,1008) = 4.05 (p= 0.00)
R = -.06 Tau-b = -.01 Rao-Scott-LR: F(24,1008) = 3.96 (p= 0.00)
Somers’ d* = -.01 Tau-c = -.01 Chisq-P(24) = 98.70
Chisq-LR(24) = 96.55
*Row variable treated as the dependent variable.

Basically consistent, but I miss the large sample size. It occurs to me that older people might be likely to have higher incomes and fewer partners. I can run more tables with restricted age ranges. I’d prefer that Vassar give age numbers he’d find acceptable, but if he’s not interested I’m open to suggestions from the comments.

UPDATE: Vassar hasn’t responded yet, but I did some thinking about the problems with the income categories. The GSS also has an occupational prestige category. I restricted it to males (SEX(1)) of AGE(18-35) and doubled the precision of the buckets as they moved outward from the center of the prestige distribution as follows PRESTG80(r: 17-24 “10th %ile”; 25-32 “10-30”; 33-47 “30-70”; 48-60 “70-90”; 61-86 “90-100”). There are a lot of prestige categories with few members, so while the percentiles are not exact they are pretty close.

Frequency Distribution
Cells contain:
-Column percent
-Weighted N
PRESTG80
1
10th %ile
2
10-30
3
30-70
4
70-90
5
90-100
ROW
TOTAL
PARTNERS 0: NONE 17.9
68.9
10.0
85.1
8.8
132.4
6.8
54.7
13.8
53.8
10.1
394.9
1: 1 50.6
194.8
52.2
441.8
64.9
974.5
64.4
516.2
66.6
260.0
60.8
2,387.3
2: 2 10.3
39.8
12.4
105.4
10.4
156.9
10.5
84.1
6.7
26.2
10.5
412.5
3: 3 7.5
29.1
8.5
72.3
6.1
92.3
7.5
59.7
5.9
23.2
7.0
276.6
4: 4 4.2
16.2
5.8
49.4
3.5
51.9
3.8
30.3
3.7
14.5
4.1
162.2
5: 5-10 5.9
22.9
7.7
65.5
4.6
68.6
6.0
47.8
2.1
8.2
5.4
213.1
6: 11-20 3.0
11.7
2.2
18.8
1.0
14.7
.7
6.0
.5
2.1
1.4
53.2
7: 21-100 .1
.6
.4
3.8
.7
10.1
.3
2.7
.7
2.6
.5
19.8
8: MT 100 .3
1.1
.5
4.5
.1
1.0
.0
.0
.0
.0
.2
6.6
COL TOTAL 100.0
385.0
100.0
846.6
100.0
1,502.5
100.0
801.5
100.0
390.6
100.0
3,926.2
Means 1.62 1.85 1.52 1.60 1.31 1.60
Std Devs 1.58 1.61 1.32 1.31 1.18 1.41
Unweighted N 311 718 1,330 764 417 3,540
Summary Statistics
Eta* = .11 Gamma = -.07 Rao-Scott-P: F(32,4192) = 3.35 (p= 0.00)
R = -.08 Tau-b = -.05 Rao-Scott-LR: F(32,4192) = 3.27 (p= 0.00)
Somers’ d* = -.04 Tau-c = -.04 Chisq-P(32) = 122.82
Chisq-LR(32) = 120.02
*Row variable treated as the dependent variable.

The pattern looks less obvious but still a trend toward less promiscuity. I see Chip recommends also looking at the oldest cohort (how old?), but I don’t feel like adding up their prestige percentiles tonight.

UPDATE 2: Jason Malloy points out a continuous variable reflectin real income, REALRINC. A scatter-plot with a best-fit line might be preferable, but you’ll have to settle for bar-charts.


UPDATE 3: Aaron requested I use PARTNRS5.

Advertisements