[Claire McKenna]

.
. // w/o fixed effects
. logit reemp3 i.hid##i.postperiod_nc b3.age_group b1.race_wbho b4.edu4 i.woman##i.marstdum1##i.ownkidd_18 b1.ind_nilf b1.
> uh_occmaj_b2 sampjl b1.durg ur_sa ur2_sa ur3_sa iur iur2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate
> _jhu stringd if cutoff3==1 & sampall==1 & age>=18 & age<65 [pw=wtfinl], vce(cluster statefip) or

note: 11.uh_occmaj_b2 omitted because of collinearity.
Iteration 0: log pseudolikelihood = -4548831
Iteration 1: log pseudolikelihood = -4170013
Iteration 2: log pseudolikelihood = -4157356.2
Iteration 3: log pseudolikelihood = -4157310.4
Iteration 4: log pseudolikelihood = -4157310.4

Logistic regression Number of obs = 2,956
Wald chi2(18) = .
Prob > chi2 = .
Log pseudolikelihood = -4157310.4 Pseudo R2 = 0.0861

(Std. err. adjusted for 19 clusters in statefip)
--------------------------------------------------------------------------------------------------------------------
| Robust
reemp3 | Odds ratio std. err. z P>|z| [95% conf. interval]
---------------------------------------------------+----------------------------------------------------------------
1.hid | .7002702 .114058 -2.19 0.029 .5088906 .9636225
1.postperiod_nc | 1.085203 .14369 0.62 0.537 .8371528 1.406751
|
hid#postperiod_nc |
1 1 | 1.76205 .2849347 3.50 0.000 1.283434 2.419151
|
age_group |
18-24 | 1.189933 .2482025 0.83 0.404 .7906317 1.790897
25-34 | .9796921 .1556727 -0.13 0.897 .71752 1.337658
45-54 | .7759897 .1527095 -1.29 0.197 .5276492 1.141213
55-64 | .7718687 .0945646 -2.11 0.035 .607099 .9813577
|
race_wbho |
2 black nh | .7084502 .2038549 -1.20 0.231 .4030677 1.245204
3 hispanic/latino | 1.117986 .2082969 0.60 0.549 .7759704 1.610747
other nh | .8385216 .1684342 -0.88 0.381 .5656309 1.24307
|
edu4 |
1 Less than HS | .8186407 .0904471 -1.81 0.070 .6592479 1.016572
2 HS or GED | .7338577 .0427831 -5.31 0.000 .6546176 .8226896
3 Some college or Associate's' | .8120404 .104228 -1.62 0.105 .631427 1.044317
|
1.woman | .8752548 .1934259 -0.60 0.547 .5675762 1.349724
1.marstdum1 | .8993397 .1351362 -0.71 0.480 .6699168 1.207332
|
woman#marstdum1 |
1 1 | .872963 .1282346 -0.92 0.355 .6545727 1.164217
|
ownkidd_18 |
1: Own children, <18, in HH | .7575139 .1594339 -1.32 0.187 .5014622 1.144308
|
woman#ownkidd_18 |
1#1: Own children, <18, in HH | 1.289486 .3991002 0.82 0.411 .7030191 2.365192
|
marstdum1#ownkidd_18 |
1#1: Own children, <18, in HH | 1.750866 .3841515 2.55 0.011 1.138921 2.69161
|
woman#marstdum1#ownkidd_18 |
1#1#1: Own children, <18, in HH | .7332693 .2613196 -0.87 0.384 .3646831 1.474387
|
ind_nilf |
2 | 1.80553 .8799177 1.21 0.225 .6946644 4.692824
3 | 1.489625 .8054717 0.74 0.461 .5161959 4.298724
4 | 1.444288 .6722382 0.79 0.430 .5800495 3.596187
5 | 1.139385 .6526263 0.23 0.820 .3707763 3.501295
6 | .7451345 .4014118 -0.55 0.585 .259229 2.141834
7 | 1.026434 .7175636 0.04 0.970 .2607773 4.040102
8 | 1.331666 .8172223 0.47 0.641 .3999691 4.433678
9 | 1.314533 .6161779 0.58 0.560 .524538 3.294322
10 | 1.808758 .920045 1.17 0.244 .6674326 4.901778
11 | 1.227343 .6951168 0.36 0.718 .4044647 3.72436
12 | 1.08146 .5506855 0.15 0.878 .3986332 2.933912
13 | 2.524114 1.81093 1.29 0.197 .6186045 10.29923
14 | 2.682423 3.342496 0.79 0.428 .2332746 30.84515
|
uh_occmaj_b2 |
professional and related occupations | 1.379627 .2331624 1.90 0.057 .990616 1.921401
service occupations | 1.010886 .187832 0.06 0.954 .7023298 1.455001
sales and related occupations | 1.676742 .1986558 4.36 0.000 1.329285 2.115019
office and administrative support occupations | 1.205044 .1652741 1.36 0.174 .9209989 1.576692
farming, fishing, and forestry occupations | 5.054637 3.232456 2.53 0.011 1.443258 17.70256
construction and extraction occupations | 1.431595 .4407435 1.17 0.244 .783 2.617451
installation, maintenance, and repair occupations | 1.760672 .3638247 2.74 0.006 1.17432 2.639797
production occupations | 1.031621 .3635775 0.09 0.930 .5170455 2.058315
transportation and material moving occupations | 1.322369 .5263619 0.70 0.483 .6060893 2.885153
armed forces | 1 (omitted)
|
sampjl | 1.226923 .0954013 2.63 0.009 1.053491 1.428907
|
durg |
5-8 weeks | .6224303 .0846339 -3.49 0.000 .4768151 .8125151
9-12 weeks | .6074873 .0809467 -3.74 0.000 .4678601 .7887845
13-16 weeks | .392522 .0648597 -5.66 0.000 .2839311 .5426441
17-20 weeks | .7426054 .1475274 -1.50 0.134 .5031029 1.096123
21-26 weeks | .3930069 .0960117 -3.82 0.000 .2434733 .6343793
27-32 weeks | .3316027 .0958552 -3.82 0.000 .188176 .5843485
33-38 weeks | .6463835 .1520183 -1.86 0.064 .4076639 1.024892
39-44 weeks | .3309865 .048453 -7.55 0.000 .2484296 .4409784
45-50 weeks | .2701556 .095354 -3.71 0.000 .1352615 .5395775
51-52 weeks | .2260038 .0388606 -8.65 0.000 .1613446 .3165753
>52 weeks | .2853079 .0616351 -5.81 0.000 .1868223 .4357114
|
ur_sa | 1.3e-114 2.1e-112 -1.62 0.105 3.0e-252 5.32e+23
ur2_sa | . . 1.49 0.136 0 .
ur3_sa | 0 0 -1.48 0.140 0 .
iur | 3.75e+89 3.27e+91 2.37 0.018 2.75e+15 5.1e+163
iur2 | 0 0 -1.77 0.076 0 .
iur3 | . . 1.50 0.133 0 .
initrate | 4.69e-85 1.08e-82 -0.84 0.400 1.8e-281 1.2e+112
initrate2 | . . 0.57 0.569 0 .
initrate3 | 0 0 -0.27 0.783 0 .
empgrowth | .8336209 .2227394 -0.68 0.496 .493778 1.407361
emp2 | .5464889 .2964898 -1.11 0.265 .188701 1.582664
emp3 | 3.837721 2.121437 2.43 0.015 1.298798 11.3398
l_incrate_jhu | 1.055488 .0706493 0.81 0.420 .9257159 1.203451
stringd | .9955981 .0077031 -0.57 0.569 .9806141 1.010811
_cons | 10.81872 22.86949 1.13 0.260 .1717268 681.5756
--------------------------------------------------------------------------------------------------------------------
Note: _cons estimates baseline odds.

. margins hid, dydx(postperiod_nc) pwcompare(cimargins effects)

Pairwise comparisons of average marginal effects

Model VCE: Robust Number of obs = 2,956

Expression: Pr(reemp3), predict()
dy/dx wrt: 1.postperiod_nc

------------------------------------------------------------------
| Delta-method Unadjusted
| Margin std. err. [95% conf. interval]
-----------------+------------------------------------------------
0.postperiod_nc | (base outcome)
-----------------+------------------------------------------------
1.postperiod_nc |
hid |
0 | .0146819 .0236426 -.0316568 .0610206
1 | .1128607 .0323135 .0495273 .176194
------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base
level.

----------------------------------------------------------------------------------
| Contrast Delta-method Unadjusted Unadjusted
| dy/dx std. err. z P>|z| [95% conf. interval]
-----------------+----------------------------------------------------------------
0.postperiod_nc | (base outcome)
-----------------+----------------------------------------------------------------
1.postperiod_nc |
hid |
1 vs 0 | .0981788 .0263574 3.72 0.000 .0465191 .1498384
----------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

.
. // w/ fixed effects
.
. logit reemp3 i.hid##i.postperiod_nc b3.age_group b1.race_wbho b4.edu4 i.woman##i.marstdum1##i.ownkidd_18 b1.ind_nilf b1.uh_occmaj_b2 samp
> jl b1.durg ur_sa ur2_sa ur3_sa iur iur2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate_jhu stringd i.year_month i.statef
> ip if cutoff3==1 & sampall==1 & age>=18 & age<65 [pw=wtfinl], vce(cluster statefip) or

note: 11.uh_occmaj_b2 omitted because of collinearity.
note: 743.year_month omitted because of collinearity.
note: 56.statefip omitted because of collinearity.
Iteration 0: log pseudolikelihood = -4548831
Iteration 1: log pseudolikelihood = -4102581.7
Iteration 2: log pseudolikelihood = -4084868
Iteration 3: log pseudolikelihood = -4084825.4
Iteration 4: log pseudolikelihood = -4084825.4

Logistic regression Number of obs = 2,956
Wald chi2(17) = .
Prob > chi2 = .
Log pseudolikelihood = -4084825.4 Pseudo R2 = 0.1020

(Std. err. adjusted for 19 clusters in statefip)
--------------------------------------------------------------------------------------------------------------------
| Robust
reemp3 | Odds ratio std. err. z P>|z| [95% conf. interval]
---------------------------------------------------+----------------------------------------------------------------
1.hid | 1.363283 .853181 0.50 0.620 .3998342 4.648277
1.postperiod_nc | 1.190843 .896832 0.23 0.817 .2721503 5.21075
|
hid#postperiod_nc |
1 1 | 1.515065 .3011399 2.09 0.037 1.026227 2.236759
|
age_group |
18-24 | 1.154654 .2576464 0.64 0.519 .7456195 1.788079
25-34 | .9550559 .1663373 -0.26 0.792 .6788585 1.343626
45-54 | .7404937 .1489847 -1.49 0.135 .4991862 1.098449
55-64 | .760808 .1021689 -2.04 0.042 .584746 .9898808
|
race_wbho |
2 black nh | .6754881 .2079867 -1.27 0.203 .3694263 1.235116
3 hispanic/latino | 1.11792 .2185224 0.57 0.569 .7621217 1.639823
other nh | .8637687 .1747419 -0.72 0.469 .5810292 1.284094
|
edu4 |
1 Less than HS | .8184452 .0919283 -1.78 0.074 .6567228 1.019993
2 HS or GED | .754165 .0455528 -4.67 0.000 .6699655 .8489466
3 Some college or Associate's' | .7973805 .1008995 -1.79 0.074 .6222367 1.021823
|
1.woman | .8557501 .197819 -0.67 0.500 .5439757 1.346215
1.marstdum1 | .860779 .1430002 -0.90 0.367 .6215592 1.192067
|
woman#marstdum1 |
1 1 | .9441973 .1355173 -0.40 0.689 .7126771 1.250929
|
ownkidd_18 |
1: Own children, <18, in HH | .7535746 .1801115 -1.18 0.237 .4717165 1.203848
|
woman#ownkidd_18 |
1#1: Own children, <18, in HH | 1.305396 .4099412 0.85 0.396 .7053983 2.41574
|
marstdum1#ownkidd_18 |
1#1: Own children, <18, in HH | 1.858785 .4656817 2.47 0.013 1.137568 3.037252
|
woman#marstdum1#ownkidd_18 |
1#1#1: Own children, <18, in HH | .6876562 .237751 -1.08 0.279 .3492015 1.35415
|
ind_nilf |
2 | 1.72103 .8403619 1.11 0.266 .6609292 4.481483
3 | 1.449865 .7524491 0.72 0.474 .5242905 4.009434
4 | 1.450273 .6554781 0.82 0.411 .5980441 3.516951
5 | 1.071389 .5806708 0.13 0.899 .370351 3.099424
6 | .7484345 .4048173 -0.54 0.592 .2592694 2.16051
7 | .9208347 .6650484 -0.11 0.909 .2235754 3.79262
8 | 1.248736 .7569864 0.37 0.714 .3806013 4.097046
9 | 1.248748 .5320873 0.52 0.602 .5417271 2.878518
10 | 1.690555 .8468961 1.05 0.295 .6333062 4.512789
11 | 1.160456 .6042375 0.29 0.775 .4182318 3.219886
12 | 1.043095 .4884148 0.09 0.928 .4166396 2.611483
13 | 2.574455 1.849768 1.32 0.188 .6296358 10.52643
14 | 2.527685 3.110576 0.75 0.451 .2265853 28.19772
|
uh_occmaj_b2 |
professional and related occupations | 1.277662 .2232481 1.40 0.161 .9071608 1.799484
service occupations | .9704075 .172422 -0.17 0.866 .6850348 1.374661
sales and related occupations | 1.607134 .1877029 4.06 0.000 1.278313 2.020538
office and administrative support occupations | 1.087359 .1455379 0.63 0.531 .8364569 1.413521
farming, fishing, and forestry occupations | 4.731459 3.031178 2.43 0.015 1.347964 16.60778
construction and extraction occupations | 1.372525 .4213739 1.03 0.302 .7519625 2.505213
installation, maintenance, and repair occupations | 1.819833 .3236284 3.37 0.001 1.284277 2.57872
production occupations | .9760549 .3437679 -0.07 0.945 .489418 1.946564
transportation and material moving occupations | 1.18488 .5009526 0.40 0.688 .5173652 2.713638
armed forces | 1 (omitted)
|
sampjl | 1.235291 .0960775 2.72 0.007 1.060633 1.438711
|
durg |
5-8 weeks | .6175598 .0764235 -3.89 0.000 .4845539 .7870746
9-12 weeks | .573018 .0730965 -4.37 0.000 .4462576 .7357851
13-16 weeks | .3682188 .0563458 -6.53 0.000 .2728053 .4970032
17-20 weeks | .7157984 .1500094 -1.60 0.111 .4746844 1.079385
21-26 weeks | .3962775 .1000765 -3.67 0.000 .2415659 .6500745
27-32 weeks | .3204143 .0910676 -4.00 0.000 .1835628 .5592924
33-38 weeks | .6623046 .1438541 -1.90 0.058 .432688 1.013773
39-44 weeks | .3340678 .0477242 -7.67 0.000 .2524839 .4420135
45-50 weeks | .2608159 .0960372 -3.65 0.000 .1267378 .5367375
51-52 weeks | .2183251 .0382605 -8.68 0.000 .1548583 .3078032
>52 weeks | .2775561 .0575554 -6.18 0.000 .1848594 .4167351
|
ur_sa | 9.3e-137 1.9e-134 -1.51 0.132 0 9.23e+40
ur2_sa | . . 1.47 0.142 0 .
ur3_sa | 0 0 -1.41 0.159 0 .
iur | 4.69e+61 5.38e+63 1.24 0.216 9.06e-37 2.4e+159
iur2 | 0 0 -1.49 0.135 0 .
iur3 | . . 1.72 0.086 0 .
initrate | 1.3e-182 5.2e-180 -1.04 0.296 0 2.3e+159
initrate2 | . . 0.36 0.721 0 .
initrate3 | . . 0.15 0.884 0 .
empgrowth | .8911005 .2501308 -0.41 0.681 .5140383 1.544749
emp2 | .7129206 .4328587 -0.56 0.577 .216882 2.343467
emp3 | 2.539794 1.524825 1.55 0.121 .7829963 8.238293
l_incrate_jhu | 1.346398 .1901113 2.11 0.035 1.020902 1.775673
stringd | 1.005602 .0117884 0.48 0.634 .9827604 1.028974
|
year_month |
733 | 1.801917 .6005156 1.77 0.077 .9376901 3.462664
734 | 2.949058 1.240613 2.57 0.010 1.293001 6.726168
735 | 1.784082 .985896 1.05 0.295 .6039973 5.269806
736 | 1.585574 .8542007 0.86 0.392 .5515898 4.557815
737 | 3.09659 2.044644 1.71 0.087 .848877 11.29595
738 | 1.951316 .6909715 1.89 0.059 .9747931 3.906094
739 | 1.243383 .4608763 0.59 0.557 .6013063 2.57107
740 | .7325305 .237433 -0.96 0.337 .3880862 1.382685
741 | 1.020987 .2614713 0.08 0.935 .6180601 1.686592
742 | .5790098 .1456082 -2.17 0.030 .3536939 .9478601
743 | 1 (omitted)
|
statefip |
arkansas | .7620739 .4846789 -0.43 0.669 .2190953 2.650704
georgia | 1.993363 1.169071 1.18 0.240 .6314983 6.29217
idaho | .6017506 .2258307 -1.35 0.176 .2883829 1.255635
iowa | 1.170538 .6345026 0.29 0.771 .4045597 3.386791
mississippi | .5234015 .2161406 -1.57 0.117 .2329851 1.175823
missouri | .6861566 .080447 -3.21 0.001 .545288 .8634171
montana | 1.566844 .62883 1.12 0.263 .7135241 3.440671
nebraska | .3698216 .3209596 -1.15 0.252 .0674923 2.026424
new hampshire | .3050378 .1093237 -3.31 0.001 .1511083 .6157707
north dakota | .8408417 .2529818 -0.58 0.564 .4662465 1.516397
oklahoma | .8012217 .1718301 -1.03 0.301 .5262644 1.219836
south carolina | .4746652 .1458261 -2.43 0.015 .2599455 .8667473
south dakota | 1.380745 .4087072 1.09 0.276 .7729541 2.466455
tennessee | .7050458 .0614784 -4.01 0.000 .5942845 .8364504
texas | .9461441 .9494129 -0.06 0.956 .1323778 6.76238
utah | .4557265 .3427238 -1.04 0.296 .1043681 1.989942
west virginia | .8811689 .6934727 -0.16 0.872 .1884389 4.12048
wyoming | 1 (omitted)
|
_cons | 7.352945 32.34522 0.45 0.650 .0013246 40818.06
--------------------------------------------------------------------------------------------------------------------
Note: _cons estimates baseline odds.

. margins hid, dydx(postperiod_nc) pwcompare(cimargins effects)

Pairwise comparisons of average marginal effects

Model VCE: Robust Number of obs = 2,956

Expression: Pr(reemp3), predict()
dy/dx wrt: 1.postperiod_nc

------------------------------------------------------------------
| Delta-method Unadjusted
| Margin std. err. [95% conf. interval]
-----------------+------------------------------------------------
0.postperiod_nc | (base outcome)
-----------------+------------------------------------------------
1.postperiod_nc |
hid |
0 | . (not estimable)
1 | . (not estimable)
------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base
level.

----------------------------------------------------------------------------------
| Contrast Delta-method Unadjusted Unadjusted
| dy/dx std. err. z P>|z| [95% conf. interval]
-----------------+----------------------------------------------------------------
0.postperiod_nc | (base outcome)
-----------------+----------------------------------------------------------------
1.postperiod_nc |
hid |
1 vs 0 | . (not estimable)
----------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

.
end of do-file

[Clyde Schechter]

Thank you, but, as I said, I also need to see the complete outputs of the two -logit- commands.

[Claire McKenna]

Thanks Clyde. Isn’t my most recent (long) post just that? I tried to copy and paste the output into the window :/. What should I do differently?

[Clyde Schechter]

Isn’t my most recent (long) post just that? I tried to copy and paste the output into the window...

Well, although you completed post #16 at 9:30, when I gave my reply in #17 at 9:35, it was not showing up in my browser. The last post I could see at that point was #15. But yes, what you show in #16 is what was needed.

note: 743.year_month omitted because of collinearity.
note: 56.statefip omitted because of collinearity.

These are the signs of what is wrong. The variables year_month and statefip both ended up having to have an additional level, other than the base category, omitted due to colinearity. That makes all the -margins- undefined, and Stata is appropriately refusing to produce some meaningless statistics for you. So the question is, what are year_month and state_fip collinear with. When a time fixed effect like year_month is colinear with something, that something us usually a variable that designates a subdivision of the timeline into periods. So if one of your variables is 0 before a certain cutoff time and after that (or similar situations), then that establishes the colinearity. When a location fixed effect like state_fip is colinear, it usually means that there is some variable that distinguishes subgroups of states (e.g maybe census region, or Republican governor vs Democratic governor, or something like that.)

The names of most of your variables do not convey any meaning to me, so I can't fully guide you as to which ones are the likely culprits. But my guess is that the offending variables will turn out to be i.hid (Does an hid ever change state?) and i.postperiod_nc (in a standard DID analysis this would normally be colinear with the time variable.) These seem likely. And, evidently, you cannot meaningfully remove them from the analysis. In this situation, you can overcome the problem by adding the -noestimcheck- to the specific -margins- command you are using, and what you get will actually be OK. But do not do that more generally, because most of the things that -margins- declares as inestimable really are undefined and cannot be estimated.

[Claire McKenna]

I see. Okay, so I'm performing this analysis on a subgroup of states (identified by dummy CUTOFF3==1 in the code), and in a specific time period, January to December 2021 (with POSTPERIOD_NC==1 for July to December; and POSTPERIOD_NC==0 for January to June). HID is a dummy referring to a policy of interest in states. I guess I do still want to control for state and month fixed effects. Where would I add noestimatecheck in this line?

margins hid, dydx(postperiod_nc) pwcompare(cimargins effects)

. Many thanks...

[Clyde Schechter]

margins hid, dydx(postperiod_nc) pwcompare(cimargins effects) noestimcheck

N.B: noestimcheck, there is no -ate- in it.

[Claire McKenna]

This works; thank you so much!

[Claire McKenna]

I just wanted to follow up with this post. I have a couple questions: 1) Stata seems to be omitting one interacted combination. Why is that? Does it impact my results? 2) I could use some help interpreting the output of this particular margins command. I post my code below. I’m interested in understanding the output of the third margins command shown in the code. Some background on the variables: the outcome is the probability of reemployment (from unemployment); HID refers to a groups of states with a particular state-level policy in place (high denials; those with lower denial rates are coded 0); CUTOFF3 refers to a group of states that ended a set of federal policies early (those that didn’t are coded 0); POSTPERIOD distinguishes between the months when the federal policies were OFF in the CUTOFF3 states and the several months prior. My output is posted at the end of this post.

logit reemp3 i.hid##i.cutoff3##i.postperiod b3.age_group b1.race_wbho b4.edu4 i.woman##i.marstdum1##i.ownkidd_18 b1.ind_nilf b1.uh_occmaj_b2 sampjl b1.durg ur_sa ur2_sa ur3_sa iur iur2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate_jhu stringd i.year_month i.statefip if sampall==1 & age>=18 & age<65 [pw=wtfinl], vce(cluster statefip) or
margins i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck
margins i.hid, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck
margins i.hid#i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck

This is how I would interpret the following lines. Am I on the right track/totally off?
(1 1) vs (0 0): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 16 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were not cut off early. This difference is significant at the 0.05 level.
(1 1) vs (0 1): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 14 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were similarly cut off early. This difference is significant at the 0.05 level.

My full output:

  	 		 			. logit reemp3 i.hid##i.cutoff3##i.postperiod b3.age_group b1.race_wbho b4.edu4 i.woman##i.marstdum1##i.ownkidd_18 b1.ind_nilf 			b1.uh_occmaj_b2 sampjl 			b1.durg ur_sa u 		 		 			> r2_sa ur3_sa iur iur2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate_jhu stringd i.year_month i.statefip if 			sampall==1 & age>=18 & 			age 		 		 			> nl], vce(cluster statefip) or 		 		 			note: 1.hid#1.cutoff3 omitted because of collinearity. 		 		 			note: 11.uh_occmaj_b2 omitted because of collinearity. 		 		 			note: 739.year_month omitted because of collinearity. 		 		 			note: 55.statefip omitted because of collinearity. 		 		 			note: 56.statefip omitted because of collinearity. 		 		 			Iteration 0: log pseudolikelihood = -12250760 		 		 			Iteration 1: log pseudolikelihood = -11236533 		 		 			Iteration 2: log pseudolikelihood = -11186703 		 		 			Iteration 3: log pseudolikelihood = -11186469 		 		 			Iteration 4: log pseudolikelihood = -11186469 		 		 			Logistic regression Number of obs = 6,684 		 		 			Wald chi2(42) = . 		 		 			Prob > chi2 = . 		 		 			Log pseudolikelihood = -11186469 Pseudo R2 = 0.0869 		 		 			(Std. err. adjusted for 44 clusters in statefip) 		 		 			Robust 		 		 			reemp3 Odds ratio std. err. z P>z [95% conf. interval] 		 		 			1.hid .8448048 .4121336 -0.35 0.730 .3247148 2.197913 		 		 			1.cutoff3 .7265061 .211079 -1.10 0.271 .4110847 1.283947 		 		 			hid#cutoff3 		 		 			1 1 1 (omitted) 		 		 			1.postperiod 1.511941 .6340878 0.99 0.324 .6645881 3.439675 		 		 			hid#postperiod 		 		 			1 1 1.319888 .2351078 1.56 0.119 .930926 1.871368 		 		 			cutoff3#postperiod 		 		 			1 1 1.173244 .2602989 0.72 0.471 .7595192 1.812334 		 		 			hid#cutoff3#postperiod 		 		 			1 1 1 1.677616 .5304469 1.64 0.102 .9027147 3.117703 		 		 			age_group 		 		 			18-24 1.067585 .1879843 0.37 0.710 .7559964 1.507597 		 		 			25-34 .9448556 .142216 -0.38 0.706 .7034705 1.269068 		 		 			45-54 .9373763 .0932369 -0.65 0.516 .7713447 1.139146 		 		 			55-64 .806711 .087214 -1.99 0.047 .6526707 .9971072 		 		 			race_wbho 		 		 			2 black nh .590971 .0838122 -3.71 0.000 .4475563 .7803413 		 		 			3 hispanic/latino .9595384 .0805638 -0.49 0.623 .8139441 1.131176 		 		 			other nh .8154017 .1073911 -1.55 0.121 .6298912 1.055547 		 		 			edu4 		 		 			1 Less than HS .9334952 .188299 -0.34 0.733 .6286557 1.386153 		 		 			2 HS or GED .7978529 .0906706 -1.99 0.047 .6385422 .9969104 		 		 			3 Some college or Associate's' .8422417 .0899708 -1.61 0.108 .6831386 1.0384 		 		 			1.woman .9602018 .0952203 -0.41 0.682 .7905902 1.166202 		 		 			1.marstdum1 1.080298 .1285047 0.65 0.516 .855639 1.363944 		 		 			woman#marstdum1 		 		 			1 1 .808204 .1726155 -1.00 0.319 .5317679 1.228344 		 		 			ownkidd_18 		 		 			1: Own children, 		 		 			woman#ownkidd_18 		 		 			1#1: Own children, 		 		 			marstdum1#ownkidd_18 		 		 			1#1: Own children, 		 		 			woman#marstdum1#ownkidd_18 		 		 			1#1#1: Own children, 		 		 			ind_nilf 		 		 			2 1.895673 .7905257 1.53 0.125 .8371418 4.292675 		 		 			3 1.046396 .4457803 0.11 0.915 .4540166 2.411685 		 		 			4 1.190101 .4815363 0.43 0.667 .5384837 2.63024 		 		 			5 1.160612 .4403797 0.39 0.695 .5517083 2.441546 		 		 			6 1.010318 .372636 0.03 0.978 .490353 2.081647 		 		 			7 1.007307 .3724497 0.02 0.984 .4880139 2.079179 		 		 			8 1.737972 .6228211 1.54 0.123 .8610048 3.508166 		 		 			9 1.380655 .5524645 0.81 0.420 .6302023 3.024758 		 		 			10 1.63282 .6146132 1.30 0.193 .7807924 3.414608 		 		 			11 1.296402 .486074 0.69 0.489 .621712 2.703273 		 		 			12 1.569633 .7231219 0.98 0.328 .6362886 3.872062 		 		 			13 1.091652 .5453244 0.18 0.861 .4100854 2.905988 		 		 			14 1.730688 1.136671 0.84 0.404 .477718 6.269975 		 		 			uh_occmaj_b2 		 		 			professional and related occupations 1.211429 .1589085 1.46 0.144 .9367891 1.566584 		 		 			service occupations 1.099327 .1602891 0.65 0.516 .8260685 1.462977 		 		 			sales and related occupations 1.312746 .222853 1.60 0.109 .9411959 1.830971 		 		 			office and administrative support occupations 1.04943 .1559791 0.32 0.745 .7842194 1.404331 		 		 			farming, fishing, and forestry occupations .953463 .5997131 -0.08 0.940 .2779148 3.271117 		 		 			construction and extraction occupations 1.693002 .2782661 3.20 0.001 1.226738 2.336485 		 		 			installation, maintenance, and repair occupations 1.411013 .2927323 1.66 0.097 .9395905 2.118963 		 		 			production occupations 1.213486 .1552048 1.51 0.130 .9444219 1.559206 		 		 			transportation and material moving occupations 1.195174 .1798934 1.18 0.236 .889839 1.605281 		 		 			armed forces 1 (omitted) 		 		 			sampjl 1.392052 .0905271 5.09 0.000 1.225464 1.581286 		 		 			durg 		 		 			5-8 weeks .6541102 .0665139 -4.17 0.000 .5359146 .7983739 		 		 			9-12 weeks .5754466 .0534491 -5.95 0.000 .4796706 .6903463 		 		 			13-16 weeks .3707656 .0485167 -7.58 0.000 .2868898 .4791635 		 		 			17-20 weeks .4360954 .0899508 -4.02 0.000 .2910779 .6533617 		 		 			21-26 weeks .4251378 .0552676 -6.58 0.000 .3295143 .5485109 		 		 			27-32 weeks .241331 .0542037 -6.33 0.000 .1553926 .3747967 		 		 			33-38 weeks .3811776 .0873856 -4.21 0.000 .2432137 .597402 		 		 			39-44 weeks .3070426 .0470368 -7.71 0.000 .2274053 .4145689 		 		 			45-50 weeks .1823773 .0407861 -7.61 0.000 .1176553 .2827028 		 		 			51-52 weeks .3386499 .0439951 -8.33 0.000 .2625238 .4368509 		 		 			>52 weeks .1587225 .0295763 -9.88 0.000 .1101605 .228692 		 		 			ur_sa 1.9e+165 2.8e+167 2.65 0.008 1.15e+43 3.2e+287 		 		 			ur2_sa 0 0 -2.57 0.010 0 0 		 		 			ur3_sa . . 2.44 0.015 . . 		 		 			iur 9.44e+18 4.74e+20 0.87 0.384 1.71e-24 5.20e+61 		 		 			iur2 4.5e-276 6.5e-273 -0.44 0.659 0 . 		 		 			iur3 . . 0.29 0.772 0 . 		 		 			initrate 2.68e-22 3.26e-20 -0.41 0.683 1.0e-125 6.95e+81 		 		 			initrate2 3.8e+281 4.2e+285 0.06 0.953 0 . 		 		 			initrate3 0 0 -0.05 0.957 0 . 		 		 			empgrowth 1.303027 .2702055 1.28 0.202 .867844 1.956433 		 		 			emp2 .9611669 .3232344 -0.12 0.906 .497216 1.858029 		 		 			emp3 1.012619 .1594366 0.08 0.937 .7437456 1.378693 		 		 			l_incrate_jhu 1.340004 .1092554 3.59 0.000 1.142101 1.5722 		 		 			stringd .9964543 .0074464 -0.48 0.635 .981966 1.011156 		 		 			year_month 		 		 			733 1.321321 .2976673 1.24 0.216 .8496688 2.054788 		 		 			734 1.900181 .5187354 2.35 0.019 1.112816 3.244642 		 		 			735 1.609576 .4028189 1.90 0.057 .9855675 2.628673 		 		 			736 1.976637 .5544794 2.43 0.015 1.140645 3.425335 		 		 			737 2.387076 1.06694 1.95 0.052 .9940514 5.732232 		 		 			738 1.110973 .1614723 0.72 0.469 .835578 1.477133 		 		 			739 1 (omitted) 		 		 			statefip 		 		 			arkansas .4891734 .222422 -1.57 0.116 .200646 1.192601 		 		 			california .5537214 .7141911 -0.46 0.647 .0441987 6.93702 		 		 			colorado .3490437 .3480586 -1.06 0.291 .0494402 2.464218 		 		 			connecticut .2703205 .3059339 -1.16 0.248 .0294128 2.484404 		 		 			delaware .6256973 .5103097 -0.57 0.565 .1265134 3.094511 		 		 			district of columbia .3567543 .4012229 -0.92 0.359 .0393611 3.233487 		 		 			georgia 1.01599 .4386541 0.04 0.971 .4358961 2.368078 		 		 			hawaii .6497519 .3648265 -0.77 0.443 .2161772 1.952923 		 		 			idaho .6562326 .2036771 -1.36 0.175 .3571639 1.205724 		 		 			illinois .3976497 .4241809 -0.86 0.387 .0491481 3.217321 		 		 			iowa .6669601 .2764237 -0.98 0.328 .296017 1.502737 		 		 			kansas 1.312426 .3608184 0.99 0.323 .7656994 2.249527 		 		 			kentucky .5488892 .3637707 -0.91 0.365 .1497475 2.011916 		 		 			maine .5152634 .3451628 -0.99 0.322 .1386211 1.915266 		 		 			maryland .8237219 .740503 -0.22 0.829 .1414406 4.797192 		 		 			michigan .5287848 .2950051 -1.14 0.253 .177174 1.578185 		 		 			minnesota .4547937 .2129583 -1.68 0.092 .1816507 1.138654 		 		 			mississippi .5688462 .1795424 -1.79 0.074 .3064319 1.05598 		 		 			missouri .7666102 .118664 -1.72 0.086 .5660003 1.038323 		 		 			montana 1.846095 .5898796 1.92 0.055 .9868951 3.453322 		 		 			nebraska 2.996182 2.303692 1.43 0.154 .6638956 13.52186 		 		 			nevada .4294347 .5259476 -0.69 0.490 .0389398 4.735884 		 		 			new hampshire .5354148 .171342 -1.95 0.051 .2859506 1.002512 		 		 			new jersey .3215604 .3689603 -0.99 0.323 .0339304 3.047447 		 		 			new mexico .3537273 .2569232 -1.43 0.152 .085195 1.468665 		 		 			new york .4069928 .5157846 -0.71 0.478 .0339511 4.878876 		 		 			north carolina .4573715 .1901148 -1.88 0.060 .2025124 1.032967 		 		 			north dakota 1.361402 .2952237 1.42 0.155 .8900222 2.082436 		 		 			oklahoma .7737274 .1862468 -1.07 0.287 .4827165 1.240178 		 		 			oregon .4425235 .3946628 -0.91 0.361 .0770541 2.541423 		 		 			pennsylvania .4519955 .4818388 -0.74 0.456 .0559412 3.652051 		 		 			rhode island .5596404 .5997285 -0.54 0.588 .0685069 4.571764 		 		 			south carolina .6372154 .1363411 -2.11 0.035 .4189477 .9691983 		 		 			south dakota 3.508452 .8599246 5.12 0.000 2.170128 5.672125 		 		 			tennessee .870771 .0831107 -1.45 0.147 .722206 1.049897 		 		 			texas .62395 .4895824 -0.60 0.548 .1340462 2.904324 		 		 			utah 1.492642 .9082895 0.66 0.510 .4528871 4.919505 		 		 			vermont .8013101 .4048566 -0.44 0.661 .2976711 2.157071 		 		 			virginia .737892 .369215 -0.61 0.544 .2767462 1.967451 		 		 			washington .3689785 .184406 -1.99 0.046 .1385455 .9826747 		 		 			west virginia .5477684 .3421977 -0.96 0.335 .1610055 1.863602 		 		 			wisconsin 1 (omitted) 		 		 			wyoming 1 (omitted) 		 		 			_cons .000032 .0000916 -3.62 0.000 1.17e-07 .0087452 		 		 			Note: _cons estimates baseline odds. 		 		 			. margins i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck 		 		 			Pairwise comparisons of average marginal effects 		 		 			Model VCE: Robust Number of obs = 6,684 		 		 			Expression: Pr(reemp3), predict() 		 		 			dy/dx wrt: 1.postperiod 		 		 			  		 		 			Delta-method Unadjusted 		 		 			Margin std. err. [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			cutoff3 		 		 			0 .0840868 .0746463 -.0622172 .2303909 		 		 			1 .1269144 .0615985 .0061836 .2476452 		 		 			Note: dy/dx for factor levels is the discrete change from the 		 		 			base level. 		 		 			  		 		 			Contrast Delta-method Unadjusted Unadjusted 		 		 			dy/dx std. err. z P>z [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			cutoff3 		 		 			1 vs 0 .0428275 .0448533 0.95 0.340 -.0450834 .1307384 		 		 			Note: dy/dx for factor levels is the discrete change from the base level. 		 		 			. margins i.hid, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck 		 		 			Pairwise comparisons of average marginal effects 		 		 			Model VCE: Robust Number of obs = 6,684 		 		 			Expression: Pr(reemp3), predict() 		 		 			dy/dx wrt: 1.postperiod 		 		 			  		 		 			Delta-method Unadjusted 		 		 			Margin std. err. [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			hid 		 		 			0 .0782285 .0719283 -.0627484 .2192055 		 		 			1 .1522654 .0583787 .0378452 .2666857 		 		 			Note: dy/dx for factor levels is the discrete change from the 		 		 			base level. 		 		 			  		 		 			Contrast Delta-method Unadjusted Unadjusted 		 		 			dy/dx std. err. z P>z [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			hid 		 		 			1 vs 0 .0740369 .0331757 2.23 0.026 .0090138 .1390601 		 		 			Note: dy/dx for factor levels is the discrete change from the base level. 		 		 			. margins i.hid#i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck 		 		 			Pairwise comparisons of average marginal effects 		 		 			Model VCE: Robust Number of obs = 6,684 		 		 			Expression: Pr(reemp3), predict() 		 		 			dy/dx wrt: 1.postperiod 		 		 			  		 		 			Delta-method Unadjusted 		 		 			Margin std. err. [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			hid#cutoff3 		 		 			0 0 .0719679 .0780209 -.0809503 .2248861 		 		 			0 1 .0903833 .0638242 -.0347098 .2154764 		 		 			1 0 .1186925 .0649391 -.0085858 .2459708 		 		 			1 1 .2344565 .0686209 .099962 .3689511 		 		 			Note: dy/dx for factor levels is the discrete change from the 		 		 			base level. 		 		 			  		 		 			Contrast Delta-method Unadjusted Unadjusted 		 		 			dy/dx std. err. z P>z [95% conf. interval] 		 		 			0.postperiod (base outcome) 		 		 			1.postperiod 		 		 			hid#cutoff3 		 		 			(0 1) vs (0 0) .0184154 .0466793 0.39 0.693 -.0730743 .109905 		 		 			(1 0) vs (0 0) .0467246 .0372942 1.25 0.210 -.0263707 .1198199 		 		 			(1 1) vs (0 0) .1624886 .0799795 2.03 0.042 .0057318 .3192455 		 		 			(1 0) vs (0 1) .0283093 .0412022 0.69 0.492 -.0524456 .1090641 		 		 			(1 1) vs (0 1) .1440733 .0575853 2.50 0.012 .0312083 .2569383 		 		 			(1 1) vs (1 0) .115764 .0688498 1.68 0.093 -.0191791 .2507072 		 		 			Note: dy/dx for factor levels is the discrete change from the base level.

[Clyde Schechter]

The output part of the post got messed up somehow, and it is all in one long line and pretty close to impossible to work with. Can you please edit or repost? Thanks.

[Claire McKenna]

Sorry about that; here's a repost (I can't edit the original post for some reason). Please let me know if this isn't legible either:
//
I just wanted to follow up with this post. I have a couple questions: 1) Stata seems to be omitting one interacted combination. Why is that? Does it impact my results? 2) I could use some help interpreting the output of this particular margins command. I post my code below. I’m interested in understanding the output of the third margins command shown in the code. Some background on the variables: the outcome is the probability of reemployment (from unemployment); HID refers to a groups of states with a particular state-level policy in place (high denials; those with lower denial rates are coded 0); CUTOFF3 refers to a group of states that ended a set of federal policies early (those that didn’t are coded 0); POSTPERIOD distinguishes between the months when the federal policies were OFF in the CUTOFF3 states and the several months prior. My output is posted at the end of this post.

logit reemp3 i.hid##i.cutoff3##i.postperiod b3.age_group b1.race_wbho b4.edu4 i.woman##i.marstdum1##i.ownkidd_18 b1.ind_nilf b1.uh_occmaj_b2 sampjl b1.durg ur_sa ur2_sa ur3_sa iur iur2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate_jhu stringd i.year_month i.statefip if sampall==1 & age>=18 & age<65 [pw=wtfinl], vce(cluster statefip) or
margins i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck
margins i.hid, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck
margins i.hid#i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck

This is how I would interpret the following lines. Am I on the right track/totally off?
(1 1) vs (0 0): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 16 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were not cut off early. This difference is significant at the 0.05 level.
(1 1) vs (0 1): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 14 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were similarly cut off early. This difference is significant at the 0.05 level.

My full output:

. logit reemp3 i.hid##i.cutoff3##i.postperiod b3.age_group b1.race_wbho b4.edu4 i.woman##i.ma
> rstdum1##i.ownkidd_18 b1.ind_nilf b1.uh_occmaj_b2 sampjl b1.durg ur_sa ur2_sa ur3_sa iur iu
> r2 iur3 initrate initrate2 initrate3 empgrowth emp2 emp3 l_incrate_jhu stringd i.year_month
> i.statefip if sampall==1 & age>=18 & age<65 [pw=wtfinl], vce(cluster statefip) or

note: 1.hid#1.cutoff3 omitted because of collinearity.
note: 11.uh_occmaj_b2 omitted because of collinearity.
note: 739.year_month omitted because of collinearity.
note: 55.statefip omitted because of collinearity.
note: 56.statefip omitted because of collinearity.
Iteration 0: log pseudolikelihood = -12250760
Iteration 1: log pseudolikelihood = -11236533
Iteration 2: log pseudolikelihood = -11186703
Iteration 3: log pseudolikelihood = -11186469
Iteration 4: log pseudolikelihood = -11186469

Logistic regression Number of obs = 6,684
Wald chi2(42) = .
Prob > chi2 = .
Log pseudolikelihood = -11186469 Pseudo R2 = 0.0869

(Std. err. adjusted for 44 clusters in statefip)
--------------------------------------------------------------------------------------------------------------------
| Robust
reemp3 | Odds ratio std. err. z P>|z| [95% conf. interval]
---------------------------------------------------+----------------------------------------------------------------
1.hid | .8448048 .4121336 -0.35 0.730 .3247148 2.197913
1.cutoff3 | .7265061 .211079 -1.10 0.271 .4110847 1.283947
|
hid#cutoff3 |
1 1 | 1 (omitted)
|
1.postperiod | 1.511941 .6340878 0.99 0.324 .6645881 3.439675
|
hid#postperiod |
1 1 | 1.319888 .2351078 1.56 0.119 .930926 1.871368
|
cutoff3#postperiod |
1 1 | 1.173244 .2602989 0.72 0.471 .7595192 1.812334
|
hid#cutoff3#postperiod |
1 1 1 | 1.677616 .5304469 1.64 0.102 .9027147 3.117703
|
age_group |
18-24 | 1.067585 .1879843 0.37 0.710 .7559964 1.507597
25-34 | .9448556 .142216 -0.38 0.706 .7034705 1.269068
45-54 | .9373763 .0932369 -0.65 0.516 .7713447 1.139146
55-64 | .806711 .087214 -1.99 0.047 .6526707 .9971072
|
race_wbho |
2 black nh | .590971 .0838122 -3.71 0.000 .4475563 .7803413
3 hispanic/latino | .9595384 .0805638 -0.49 0.623 .8139441 1.131176
other nh | .8154017 .1073911 -1.55 0.121 .6298912 1.055547
|
edu4 |
1 Less than HS | .9334952 .188299 -0.34 0.733 .6286557 1.386153
2 HS or GED | .7978529 .0906706 -1.99 0.047 .6385422 .9969104
3 Some college or Associate's' | .8422417 .0899708 -1.61 0.108 .6831386 1.0384
|
1.woman | .9602018 .0952203 -0.41 0.682 .7905902 1.166202
1.marstdum1 | 1.080298 .1285047 0.65 0.516 .855639 1.363944
|
woman#marstdum1 |
1 1 | .808204 .1726155 -1.00 0.319 .5317679 1.228344
|
ownkidd_18 |
1: Own children, <18, in HH | .874251 .1896057 -0.62 0.535 .5715173 1.337343
|
woman#ownkidd_18 |
1#1: Own children, <18, in HH | 1.048349 .2681951 0.18 0.854 .6349603 1.730874
|
marstdum1#ownkidd_18 |
1#1: Own children, <18, in HH | 1.397039 .2992478 1.56 0.119 .9180782 2.125872
|
woman#marstdum1#ownkidd_18 |
1#1#1: Own children, <18, in HH | .8722012 .3426738 -0.35 0.728 .403825 1.883823
|
ind_nilf |
2 | 1.895673 .7905257 1.53 0.125 .8371418 4.292675
3 | 1.046396 .4457803 0.11 0.915 .4540166 2.411685
4 | 1.190101 .4815363 0.43 0.667 .5384837 2.63024
5 | 1.160612 .4403797 0.39 0.695 .5517083 2.441546
6 | 1.010318 .372636 0.03 0.978 .490353 2.081647
7 | 1.007307 .3724497 0.02 0.984 .4880139 2.079179
8 | 1.737972 .6228211 1.54 0.123 .8610048 3.508166
9 | 1.380655 .5524645 0.81 0.420 .6302023 3.024758
10 | 1.63282 .6146132 1.30 0.193 .7807924 3.414608
11 | 1.296402 .486074 0.69 0.489 .621712 2.703273
12 | 1.569633 .7231219 0.98 0.328 .6362886 3.872062
13 | 1.091652 .5453244 0.18 0.861 .4100854 2.905988
14 | 1.730688 1.136671 0.84 0.404 .477718 6.269975
|
uh_occmaj_b2 |
professional and related occupations | 1.211429 .1589085 1.46 0.144 .9367891 1.566584
service occupations | 1.099327 .1602891 0.65 0.516 .8260685 1.462977
sales and related occupations | 1.312746 .222853 1.60 0.109 .9411959 1.830971
office and administrative support occupations | 1.04943 .1559791 0.32 0.745 .7842194 1.404331
farming, fishing, and forestry occupations | .953463 .5997131 -0.08 0.940 .2779148 3.271117
construction and extraction occupations | 1.693002 .2782661 3.20 0.001 1.226738 2.336485
installation, maintenance, and repair occupations | 1.411013 .2927323 1.66 0.097 .9395905 2.118963
production occupations | 1.213486 .1552048 1.51 0.130 .9444219 1.559206
transportation and material moving occupations | 1.195174 .1798934 1.18 0.236 .889839 1.605281
armed forces | 1 (omitted)
|
sampjl | 1.392052 .0905271 5.09 0.000 1.225464 1.581286
|
durg |
5-8 weeks | .6541102 .0665139 -4.17 0.000 .5359146 .7983739
9-12 weeks | .5754466 .0534491 -5.95 0.000 .4796706 .6903463
13-16 weeks | .3707656 .0485167 -7.58 0.000 .2868898 .4791635
17-20 weeks | .4360954 .0899508 -4.02 0.000 .2910779 .6533617
21-26 weeks | .4251378 .0552676 -6.58 0.000 .3295143 .5485109
27-32 weeks | .241331 .0542037 -6.33 0.000 .1553926 .3747967
33-38 weeks | .3811776 .0873856 -4.21 0.000 .2432137 .597402
39-44 weeks | .3070426 .0470368 -7.71 0.000 .2274053 .4145689
45-50 weeks | .1823773 .0407861 -7.61 0.000 .1176553 .2827028
51-52 weeks | .3386499 .0439951 -8.33 0.000 .2625238 .4368509
>52 weeks | .1587225 .0295763 -9.88 0.000 .1101605 .228692
|
ur_sa | 1.9e+165 2.8e+167 2.65 0.008 1.15e+43 3.2e+287
ur2_sa | 0 0 -2.57 0.010 0 0
ur3_sa | . . 2.44 0.015 . .
iur | 9.44e+18 4.74e+20 0.87 0.384 1.71e-24 5.20e+61
iur2 | 4.5e-276 6.5e-273 -0.44 0.659 0 .
iur3 | . . 0.29 0.772 0 .
initrate | 2.68e-22 3.26e-20 -0.41 0.683 1.0e-125 6.95e+81
initrate2 | 3.8e+281 4.2e+285 0.06 0.953 0 .
initrate3 | 0 0 -0.05 0.957 0 .
empgrowth | 1.303027 .2702055 1.28 0.202 .867844 1.956433
emp2 | .9611669 .3232344 -0.12 0.906 .497216 1.858029
emp3 | 1.012619 .1594366 0.08 0.937 .7437456 1.378693
l_incrate_jhu | 1.340004 .1092554 3.59 0.000 1.142101 1.5722
stringd | .9964543 .0074464 -0.48 0.635 .981966 1.011156
|
year_month |
733 | 1.321321 .2976673 1.24 0.216 .8496688 2.054788
734 | 1.900181 .5187354 2.35 0.019 1.112816 3.244642
735 | 1.609576 .4028189 1.90 0.057 .9855675 2.628673
736 | 1.976637 .5544794 2.43 0.015 1.140645 3.425335
737 | 2.387076 1.06694 1.95 0.052 .9940514 5.732232
738 | 1.110973 .1614723 0.72 0.469 .835578 1.477133
739 | 1 (omitted)
|
statefip |
arkansas | .4891734 .222422 -1.57 0.116 .200646 1.192601
california | .5537214 .7141911 -0.46 0.647 .0441987 6.93702
colorado | .3490437 .3480586 -1.06 0.291 .0494402 2.464218
connecticut | .2703205 .3059339 -1.16 0.248 .0294128 2.484404
delaware | .6256973 .5103097 -0.57 0.565 .1265134 3.094511
district of columbia | .3567543 .4012229 -0.92 0.359 .0393611 3.233487
georgia | 1.01599 .4386541 0.04 0.971 .4358961 2.368078
hawaii | .6497519 .3648265 -0.77 0.443 .2161772 1.952923
idaho | .6562326 .2036771 -1.36 0.175 .3571639 1.205724
illinois | .3976497 .4241809 -0.86 0.387 .0491481 3.217321
iowa | .6669601 .2764237 -0.98 0.328 .296017 1.502737
kansas | 1.312426 .3608184 0.99 0.323 .7656994 2.249527
kentucky | .5488892 .3637707 -0.91 0.365 .1497475 2.011916
maine | .5152634 .3451628 -0.99 0.322 .1386211 1.915266
maryland | .8237219 .740503 -0.22 0.829 .1414406 4.797192
michigan | .5287848 .2950051 -1.14 0.253 .177174 1.578185
minnesota | .4547937 .2129583 -1.68 0.092 .1816507 1.138654
mississippi | .5688462 .1795424 -1.79 0.074 .3064319 1.05598
missouri | .7666102 .118664 -1.72 0.086 .5660003 1.038323
montana | 1.846095 .5898796 1.92 0.055 .9868951 3.453322
nebraska | 2.996182 2.303692 1.43 0.154 .6638956 13.52186
nevada | .4294347 .5259476 -0.69 0.490 .0389398 4.735884
new hampshire | .5354148 .171342 -1.95 0.051 .2859506 1.002512
new jersey | .3215604 .3689603 -0.99 0.323 .0339304 3.047447
new mexico | .3537273 .2569232 -1.43 0.152 .085195 1.468665
new york | .4069928 .5157846 -0.71 0.478 .0339511 4.878876
north carolina | .4573715 .1901148 -1.88 0.060 .2025124 1.032967
north dakota | 1.361402 .2952237 1.42 0.155 .8900222 2.082436
oklahoma | .7737274 .1862468 -1.07 0.287 .4827165 1.240178
oregon | .4425235 .3946628 -0.91 0.361 .0770541 2.541423
pennsylvania | .4519955 .4818388 -0.74 0.456 .0559412 3.652051
rhode island | .5596404 .5997285 -0.54 0.588 .0685069 4.571764
south carolina | .6372154 .1363411 -2.11 0.035 .4189477 .9691983
south dakota | 3.508452 .8599246 5.12 0.000 2.170128 5.672125
tennessee | .870771 .0831107 -1.45 0.147 .722206 1.049897
texas | .62395 .4895824 -0.60 0.548 .1340462 2.904324
utah | 1.492642 .9082895 0.66 0.510 .4528871 4.919505
vermont | .8013101 .4048566 -0.44 0.661 .2976711 2.157071
virginia | .737892 .369215 -0.61 0.544 .2767462 1.967451
washington | .3689785 .184406 -1.99 0.046 .1385455 .9826747
west virginia | .5477684 .3421977 -0.96 0.335 .1610055 1.863602
wisconsin | 1 (omitted)
wyoming | 1 (omitted)
|
_cons | .000032 .0000916 -3.62 0.000 1.17e-07 .0087452
--------------------------------------------------------------------------------------------------------------------
Note: _cons estimates baseline odds.

. margins i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck

Pairwise comparisons of average marginal effects

Model VCE: Robust Number of obs = 6,684

Expression: Pr(reemp3), predict()
dy/dx wrt: 1.postperiod

---------------------------------------------------------------
| Delta-method Unadjusted
| Margin std. err. [95% conf. interval]
--------------+------------------------------------------------
0.postperiod | (base outcome)
--------------+------------------------------------------------
1.postperiod |
cutoff3 |
0 | .0840868 .0746463 -.0622172 .2303909
1 | .1269144 .0615985 .0061836 .2476452
---------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the
base level.

-------------------------------------------------------------------------------
| Contrast Delta-method Unadjusted Unadjusted
| dy/dx std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
0.postperiod | (base outcome)
--------------+----------------------------------------------------------------
1.postperiod |
cutoff3 |
1 vs 0 | .0428275 .0448533 0.95 0.340 -.0450834 .1307384
-------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

. margins i.hid, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck

Pairwise comparisons of average marginal effects

Model VCE: Robust Number of obs = 6,684

Expression: Pr(reemp3), predict()
dy/dx wrt: 1.postperiod

---------------------------------------------------------------
| Delta-method Unadjusted
| Margin std. err. [95% conf. interval]
--------------+------------------------------------------------
0.postperiod | (base outcome)
--------------+------------------------------------------------
1.postperiod |
hid |
0 | .0782285 .0719283 -.0627484 .2192055
1 | .1522654 .0583787 .0378452 .2666857
---------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the
base level.

-------------------------------------------------------------------------------
| Contrast Delta-method Unadjusted Unadjusted
| dy/dx std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
0.postperiod | (base outcome)
--------------+----------------------------------------------------------------
1.postperiod |
hid |
1 vs 0 | .0740369 .0331757 2.23 0.026 .0090138 .1390601
-------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

. margins i.hid#i.cutoff3, dydx(i.postperiod) pwcompare(cimargins effects) noestimcheck

Pairwise comparisons of average marginal effects

Model VCE: Robust Number of obs = 6,684

Expression: Pr(reemp3), predict()
dy/dx wrt: 1.postperiod

---------------------------------------------------------------
| Delta-method Unadjusted
| Margin std. err. [95% conf. interval]
--------------+------------------------------------------------
0.postperiod | (base outcome)
--------------+------------------------------------------------
1.postperiod |
hid#cutoff3 |
0 0 | .0719679 .0780209 -.0809503 .2248861
0 1 | .0903833 .0638242 -.0347098 .2154764
1 0 | .1186925 .0649391 -.0085858 .2459708
1 1 | .2344565 .0686209 .099962 .3689511
---------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the
base level.

---------------------------------------------------------------------------------
| Contrast Delta-method Unadjusted Unadjusted
| dy/dx std. err. z P>|z| [95% conf. interval]
----------------+----------------------------------------------------------------
0.postperiod | (base outcome)
----------------+----------------------------------------------------------------
1.postperiod |
hid#cutoff3 |
(0 1) vs (0 0) | .0184154 .0466793 0.39 0.693 -.0730743 .109905
(1 0) vs (0 0) | .0467246 .0372942 1.25 0.210 -.0263707 .1198199
(1 1) vs (0 0) | .1624886 .0799795 2.03 0.042 .0057318 .3192455
(1 0) vs (0 1) | .0283093 .0412022 0.69 0.492 -.0524456 .1090641
(1 1) vs (0 1) | .1440733 .0575853 2.50 0.012 .0312083 .2569383
(1 1) vs (1 0) | .115764 .0688498 1.68 0.093 -.0191791 .2507072
---------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

[Clyde Schechter]

This is how I would interpret the following lines. Am I on the right track/totally off?
(1 1) vs (0 0): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 16 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were not cut off early. This difference is significant at the 0.05 level.
(1 1) vs (0 1): In high denial states, where benefits were cut off early, the increase in the probability of reemployment in the post-period relative to the pre-period was 14 percentage points greater than the increase in the probability of reemployment in the post-period relative to the pre-period in non-high denial states, where benefits were similarly cut off early. This difference is significant at the 0.05 level.

Right on track! At least, within the framework of statistical significance testing.

In my view, which does not use the framework of statistical significance testing, one should not overlook the (1 1) vs (0 1) comparison. While it is not "statistically significant," the effect estimate is only slightly smaller than the other two, and all three confidence intervals overlap almost completely. It is true that the data are compatible with no difference, or even a tiny negative difference, for the (1 1) vs (0 1) comparison, it is also true that the data are compatible with very, very small (1% percentage point or even a fraction of a percentage point) differences in probabilty of re-employment. Labor market analysis is not my field, but to me a difference that small does not seem large enough to be worth talking about, even if it is, strictly speaking, positive. So I would not consider the (1 1) vs (0 1) contrast as being qualitatively a different beast from the other two contrasts, and I would present them in a more or less equivalent manner.

[Claire McKenna]

That's great; thanks so much. Any thoughts on the dropped interaction? HID#CUTOFF3

[Clyde Schechter]

Any thoughts on the dropped interaction? HID#CUTOFF3

Ouch! I got distracted by the other question and forgot about this.

This worries me. In particular, reading this more carefully, I see that you actually have two treatments, crossed with each other, and you have a prepost variable. But how is that prepost variable defined? Do the two treatments always begin simultaneously? If so, then there is no problem defining prepost. But if not, then it seems that a single prepost variable is not a proper specification of what's going on.

Although the above does not seem directly related to your question, my mind was drawn to it because one possible reason for an HID#CUTOFF3 interaction term to drop is if its value is automatically determined by the value of the prepost variable. For example, this would happen if prepost is specified as always 0 if one or both treatments is 0. You can easily check this by running -by prepost, sort: tab HID CUTOFF3-. If you have any 0 cells in either table, that's would be the cause of the dropped term, and I think it could indicate a mis-specified model.

[Claire McKenna]

Okay I see. Yeah, I'm trying to understand how reemployment changed pre-/post-cutoff in states that are strict on a particular measure relative to states that aren't strict on a particular measure (HID).

I just ran your suggested command and this is what came out [EDITED FROM ORIGINAL POST -- I'M ALL TURNED AROUND]:

.. by postperiod, sort: tab hid cutoff3

    
-> postperiod = 0

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1      Total

0    186,560     91,271    277,831 
1     63,792     90,134    153,926 

Total    250,352    181,405    431,757 

    
-> postperiod = 1

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1      Total

0     60,092     29,980     90,072 
1     21,134     29,155     50,289 

Total     81,226     59,135    140,361 

    
-> postperiod = .

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1      Total

0  1,072,417    520,647  1,593,064 
1    361,871    506,384    868,255 

Total  1,434,288  1,027,031  2,461,319 

[Claire McKenna]

This actually captures my sample limitations (apologies)

. by postperiod, sort: tab hid cutoff3    if sampall==1    &    age>=18    &    age<65

                        
-> postperiod = 0

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1    Total
    
0      2,783        949    3,732 
1        791        937    1,728 
    
Total      3,574      1,886    5,460 

                        
-> postperiod = 1

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1    Total
    
0        794        246    1,040 
1        270        230    500 
    
Total      1,064        476    1,540 

                        
-> postperiod = .

High NS 
denials 
states, as 
of 1/2020 
(1==high; 
0==low or         cutoff3
medium)          0          1    Total
    
0     12,454      4,823    17,277 
1      3,947      4,309    8,256 
    
Total     16,401      9,132    25,533