Robin Hanson responded to a Washington Post article (possibly inspired by Spotted Toad) on the increasing percentage of males 18-29 who report not having sex in the past year with some speculation on whether that was attributed to women of that age group (who reported a smaller increase in celibacy) shifting toward older men or to that subset of 18-29 year old men with more partners. It struck me that since the source of this data was the General Social Survey, which asks respondents their age as well as the number of partners, it should be answerable directly rather than guesses from respondents to a twitter poll. My initial attempt to do so was stymied by a newer GSS interface which generated errors when I tried to construct variables, but an anonymous commenter elsewhere pointed me toward the old interface which was still working. The parameters I used were as follows:
Row: PARTNERS
Column: AGE(r:18-29; 30-39; 40-49; 50-59)
Control(s): YEAR
Selection filter(s): SEX(1), YEAR(2008, 2010, 2012, 2014, 2016, 2018), NUMMEN(0), PARTNERS(0-8)

Variables
Role Name Label Range MD Dataset
Row PARTNERS How many sex partner’s R had in last year 0-9 -1,98,99 1
Column AGE(Recoded) Age of respondent 1-4 1
Control YEAR GSS year for this respondent 1972-2018 1
Weight COMPWT Composite weight = WTSSALL * OVERSAMP * FORMWT .1912-11.1196 1
Filter SEX(1) Respondents sex(=MALE) 1-2 0 1
Filter YEAR(2008,

2010,

2012,

2014,

2016,

2018)

GSS year for this respondent 1972-2018 1
Filter NUMMEN(0) Number of male sex partner’s since 18 0-997 -1,998,999 1
Filter PARTNERS(0-8) How many sex partner’s R had in last year 0-9 -1,98,99 1
Statistics for YEAR = 2008
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 13.1
20.8
7.5
10.3
13.1
19.0
10.5
15.3
11.2
65.3
1: 1 PARTNER 48.6
76.8
78.3
107.2
71.2
103.4
83.6
121.5
69.8
408.9
2: 2 PARTNERS 14.8
23.4
6.5
8.9
6.9
10.0
2.7
3.9
7.9
46.2
3: 3 PARTNERS 4.7
7.4
2.2
3.1
3.4
5.0
1.2
1.8
2.9
17.3
4: 4 PARTNERS 3.6
5.6
4.2
5.8
2.1
3.0
.9
1.3
2.7
15.7
5: 5-10 PARTNERS 7.2
11.5
1.3
1.8
3.3
4.8
.7
1.1
3.3
19.0
6: 11-20 PARTNERS 4.1
6.6
.0
.0
.0
.0
.3
.4
1.2
7.0
7: 21-100 PARTNERS 2.8
4.4
.0
.0
.0
.0
.0
.0
.7
4.4
8: MORE THAN 100 PARTNERS 1.1
1.8
.0
.0
.0
.0
.0
.0
.3
1.8
COL TOTAL 100.0
158.2
100.0
136.9
100.0
145.2
100.0
145.3
100.0
585.6
Means 1.96 1.21 1.20 1.02 1.36
Std Devs 1.90 .88 1.01 .68 1.29
Unweighted N 133 128 146 135 542
Summary Statistics for YEAR = 2008
Eta* = .29 Gamma = -.30 Rao-Scott-P: F(24,3696) = 4.61 (p= 0.00)
R = -.26 Tau-b = -.18 Rao-Scott-LR: F(24,3696) = 4.61 (p= 0.00)
Somers’ d* = -.15 Tau-c = -.15 Chisq-P(24) = 78.79
Chisq-LR(24) = 78.75
*Row variable treated as the dependent variable.
Statistics for YEAR = 2010
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 11.9
19.4
5.9
8.0
5.9
8.2
13.6
18.0
9.4
53.7
1: 1 PARTNER 52.0
84.9
75.8
103.2
82.6
116.2
77.3
102.2
71.0
406.6
2: 2 PARTNERS 10.3
16.8
4.6
6.3
6.1
8.6
6.3
8.4
7.0
40.1
3: 3 PARTNERS 15.6
25.5
3.6
4.9
2.8
4.0
1.7
2.3
6.4
36.7
4: 4 PARTNERS 7.4
12.1
2.3
3.2
2.2
3.2
.3
.5
3.3
18.9
5: 5-10 PARTNERS 2.3
3.7
6.8
9.2
.3
.5
.7
.9
2.5
14.3
6: 11-20 PARTNERS .6
.9
1.0
1.4
.0
.0
.0
.0
.4
2.3
COL TOTAL 100.0
163.4
100.0
136.2
100.0
140.7
100.0
132.3
100.0
572.5
Means 1.64 1.45 1.14 1.00 1.32
Std Devs 1.27 1.27 .69 .64 1.06
Unweighted N 134 125 137 132 528
Summary Statistics for YEAR = 2010
Eta* = .24 Gamma = -.31 Rao-Scott-P: F(18,2772) = 5.07 (p= 0.00)
R = -.24 Tau-b = -.19 Rao-Scott-LR: F(18,2772) = 4.91 (p= 0.00)
Somers’ d* = -.15 Tau-c = -.15 Chisq-P(18) = 76.73
Chisq-LR(18) = 74.26
*Row variable treated as the dependent variable.
Statistics for YEAR = 2012
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 20.4
32.3
12.4
18.7
5.6
6.7
16.1
20.1
14.1
77.8
1: 1 PARTNER 47.6
75.4
70.8
106.5
83.6
99.0
77.4
96.8
68.4
377.6
2: 2 PARTNERS 12.4
19.6
8.3
12.6
4.6
5.5
4.4
5.6
7.8
43.2
3: 3 PARTNERS 7.0
11.1
4.0
6.0
2.1
2.5
.7
.8
3.7
20.4
4: 4 PARTNERS 6.3
9.9
.8
1.2
3.3
3.9
1.4
1.7
3.0
16.8
5: 5-10 PARTNERS 5.2
8.3
2.2
3.3
.3
.4
.0
.0
2.2
12.0
6: 11-20 PARTNERS .6
.9
.3
.4
.3
.4
.0
.0
.3
1.7
7: 21-100 PARTNERS .5
.8
1.2
1.7
.0
.0
.0
.0
.5
2.6
COL TOTAL 100.0
158.2
100.0
150.3
100.0
118.4
100.0
125.0
100.0
551.9
Means 1.52 1.23 1.16 .94 1.23
Std Devs 1.44 1.11 .78 .59 1.09
Unweighted N 133 148 111 114 506
Summary Statistics for YEAR = 2012
Eta* = .19 Gamma = -.19 Rao-Scott-P: F(21,3234) = 3.49 (p= 0.00)
R = -.19 Tau-b = -.12 Rao-Scott-LR: F(21,3234) = 3.78 (p= 0.00)
Somers’ d* = -.10 Tau-c = -.09 Chisq-P(21) = 61.20
Chisq-LR(21) = 66.34
*Row variable treated as the dependent variable.
Statistics for YEAR = 2014
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 18.5
36.9
7.4
12.5
6.0
8.3
11.4
22.1
11.4
79.8
1: 1 PARTNER 54.8
109.0
76.3
128.7
83.2
115.5
80.4
156.0
72.7
509.1
2: 2 PARTNERS 9.8
19.6
5.6
9.5
3.6
5.0
4.3
8.3
6.1
42.4
3: 3 PARTNERS 6.8
13.6
4.3
7.2
3.4
4.8
2.8
5.4
4.4
31.0
4: 4 PARTNERS 5.0
10.0
2.9
4.9
.8
1.1
.2
.4
2.4
16.5
5: 5-10 PARTNERS 3.4
6.8
1.8
3.0
3.0
4.1
.9
1.8
2.2
15.7
6: 11-20 PARTNERS 1.6
3.1
1.2
2.1
.0
.0
.0
.0
.7
5.2
7: 21-100 PARTNERS .0
.0
.4
.7
.0
.0
.0
.0
.1
.7
COL TOTAL 100.0
199.1
100.0
168.5
100.0
138.7
100.0
194.0
100.0
700.4
Means 1.42 1.31 1.19 1.03 1.24
Std Devs 1.34 1.10 .87 .66 1.04
Unweighted N 162 165 147 187 661
Summary Statistics for YEAR = 2014
Eta* = .15 Gamma = -.14 Rao-Scott-P: F(21,3234) = 3.23 (p= 0.00)
R = -.14 Tau-b = -.08 Rao-Scott-LR: F(21,3234) = 3.39 (p= 0.00)
Somers’ d* = -.06 Tau-c = -.06 Chisq-P(21) = 57.03
Chisq-LR(21) = 59.96
*Row variable treated as the dependent variable.
Statistics for YEAR = 2016
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 23.2
37.3
10.9
15.0
14.5
16.8
13.3
15.2
16.0
84.2
1: 1 PARTNER 37.4
60.2
72.6
99.5
76.7
88.6
75.7
86.6
63.4
334.8
2: 2 PARTNERS 17.7
28.4
9.1
12.5
7.6
8.7
4.9
5.6
10.5
55.2
3: 3 PARTNERS 4.8
7.7
1.4
1.9
.8
1.0
2.9
3.3
2.6
13.9
4: 4 PARTNERS 7.5
12.0
1.8
2.5
.4
.5
1.3
1.4
3.1
16.5
5: 5-10 PARTNERS 6.8
11.0
3.1
4.3
.0
.0
1.5
1.7
3.2
17.0
6: 11-20 PARTNERS 2.7
4.3
1.0
1.4
.0
.0
.4
.5
1.2
6.2
COL TOTAL 100.0
160.9
100.0
137.1
100.0
115.5
100.0
114.4
100.0
527.9
Means 1.67 1.24 .96 1.09 1.28
Std Devs 1.61 1.07 .54 .87 1.18
Unweighted N 140 134 108 117 499
Summary Statistics for YEAR = 2016
Eta* = .24 Gamma = -.19 Rao-Scott-P: F(18,2772) = 4.59 (p= 0.00)
R = -.20 Tau-b = -.13 Rao-Scott-LR: F(18,2772) = 4.84 (p= 0.00)
Somers’ d* = -.11 Tau-c = -.11 Chisq-P(18) = 76.97
Chisq-LR(18) = 81.11
*Row variable treated as the dependent variable.
Statistics for YEAR = 2018
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 28.5
43.1
8.8
9.2
5.9
5.7
12.8
12.0
15.7
70.0
1: 1 PARTNER 39.9
60.3
71.6
74.9
82.9
79.0
79.6
74.5
64.9
288.7
2: 2 PARTNERS 11.4
17.2
7.2
7.5
6.2
5.9
3.5
3.3
7.6
33.9
3: 3 PARTNERS 10.9
16.5
2.5
2.6
2.0
1.9
3.0
2.8
5.4
23.8
4: 4 PARTNERS 5.0
7.5
2.3
2.4
2.0
1.9
.5
.5
2.8
12.3
5: 5-10 PARTNERS 4.0
6.1
7.7
8.0
1.0
.9
.5
.5
3.5
15.6
7: 21-100 PARTNERS .3
.5
.0
.0
.0
.0
.0
.0
.1
.5
COL TOTAL 100.0
151.3
100.0
104.6
100.0
95.2
100.0
93.5
100.0
444.7
Means 1.38 1.41 1.14 1.00 1.25
Std Devs 1.38 1.24 .72 .64 1.11
Unweighted N 113 103 82 93 391
Summary Statistics for YEAR = 2018
Eta* = .15 Gamma = -.07 Rao-Scott-P: F(18,2772) = 4.74 (p= 0.00)
R = -.14 Tau-b = -.05 Rao-Scott-LR: F(18,2772) = 4.84 (p= 0.00)
Somers’ d* = -.04 Tau-c = -.04 Chisq-P(18) = 72.01
Chisq-LR(18) = 73.55
*Row variable treated as the dependent variable.
Statistics for all valid cases
Cells contain:
Column percent
-Weighted N
AGE
1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
PARTNERS 0: NO PARTNERS 19.1
189.8
8.8
73.6
8.6
64.7
12.8
102.7
12.7
430.8
1: 1 PARTNER 47.1
466.7
74.4
619.9
79.8
601.6
79.3
637.6
68.7
2,325.7
2: 2 PARTNERS 12.6
125.0
6.9
57.3
5.8
43.7
4.4
35.0
7.7
261.1
3: 3 PARTNERS 8.2
81.8
3.1
25.7
2.5
19.1
2.0
16.5
4.2
143.0
4: 4 PARTNERS 5.8
57.2
2.4
20.0
1.8
13.6
.7
5.8
2.9
96.6
5: 5-10 PARTNERS 4.8
47.4
3.5
29.5
1.4
10.7
.7
6.0
2.8
93.6
6: 11-20 PARTNERS 1.6
15.8
.6
5.3
.1
.4
.1
.9
.7
22.4
7: 21-100 PARTNERS .6
5.7
.3
2.4
.0
.0
.0
.0
.2
8.1
8: MORE THAN 100 PARTNERS .2
1.8
.0
.0
.0
.0
.0
.0
.1
1.8
COL TOTAL 100.0
991.1
100.0
833.7
100.0
753.8
100.0
804.5
100.0
3,383.0
Means 1.59 1.30 1.14 1.01 1.28
Std Devs 1.51 1.11 .79 .68 1.13
Unweighted N 815 803 731 778 3,127
Summary Statistics for all valid cases
Eta* = .20 Gamma = -.20 Rao-Scott-P: F(24,3696) = 15.19 (p= 0.00)
R = -.19 Tau-b = -.12 Rao-Scott-LR: F(24,3696) = 15.35 (p= 0.00)
Somers’ d* = -.10 Tau-c = -.10 Chisq-P(24) = 336.58
Chisq-LR(24) = 340.18
*Row variable treated as the dependent variable.

Tracking the standard deviations for the first age group (18-29) by survey year we have

Year Standard deviations
2008 1.90
2010 1.27
2012 1.44
2014 1.34
2016 1.61
2018 1.38
All Years 1.51

2008 actually had the highest standard deviation, and while 2010 had the lowest there wasn’t anything like a consistent pattern of increase, as 2018 had the second lowest. Now let’s mean at the means for all age groups per year:

 

Year 1
18-29
2
30-39
3
40-49
4
50-59
ROW
TOTAL
2008 1.96 1.21 1.20 1.02 1.36
2010 1.64 1.45 1.14 1.00 1.32
2012 1.52 1.23 1.16 0.94 1.23
2014 1.42 1.31 1.19 1.03 1.24
2016 1.67 1.24 0.96 1.09 1.28
2018 1.38 1.41 1.14 1.00 1.25
All Years 1.59 1.30 1.14 1.01 1.28

The overall mean trends downward, but the 50-59 group actually had their mean increase. The 18-29 group had a significant downward trend, even with the blip upward in 2016. 2016 is also responsible for most of the difference between the first & last half of the years in the 40-49 group. Between the two explanations, I would thus conclude that a shift toward older men relative to the 18-29 group fits better than an increase in variance within 18-29. A major caveat: the PARTNERS variable already comes in buckets starting with 5, so you can’t really measure aggregates in terms of actual numbers of partners. I kept those as still preserving rank ordering and contributing to shifts in means & variance, even if it wouldn’t do justice to a hypothetical where those men are taking up a large fraction of all partnerings. Also, the more important an individual man might be, the bigger an issue the sample size presents.