============================================================================
A. SHOPPING x SELF-EMPLOYMENT
============================================================================
Users n=5783, self-employed (business module) = 396 (6.8%)

[A1] use_shop separately by employment type:
  Self-employed users (n=396):
    use_shop           b=+0.1898 (SE=0.1158, p=0.101)
  Wage-employed users (n=5387):
    use_shop           b=+0.0634*** (SE=0.0231, p=0.006)

[A2] pooled interaction (use_shop x selfemp):
  Pooled with interaction (n=5783):
    use_shop           b=+0.0688*** (SE=0.0229, p=0.003)
    selfemp            b=-0.5589*** (SE=0.0857, p=0.000)
    shop_x_self        b=+0.0815 (SE=0.1176, p=0.488)

[A3] same split for use_learn (placebo/contrast):
  Self-employed users (n=396):
    use_learn          b=+0.1602 (SE=0.1492, p=0.283)
  Wage-employed users (n=5387):
    use_learn          b=+0.0573** (SE=0.0248, p=0.021)

============================================================================
B. SWITCHERS vs NON-SWITCHERS  [4-wave panel, incomeb]
============================================================================
                age_  gender_  eduyear  urban  married  ln_income     N
group                                                                  
always-online  31.50     0.56    12.23   0.68     0.68      10.07  4878
never-online   45.98     0.64     6.90   0.52     0.92       9.81  1042
switcher       37.54     0.63     9.44   0.55     0.83       9.94  2168

(first-observed-wave characteristics; switchers sit between never- and always-online groups)

============================================================================
C. USE-TYPE CORRELATIONS among users
============================================================================
           use_learn  use_shop  use_game
use_learn      1.000     0.129     0.050
use_shop       0.129     1.000     0.067
use_game       0.050     0.067     1.000

Overlap: learn&shop 19.6% | learn&game 8.3% | shop&game 17.0% | all three 5.9%

============================================================================
D. SAMPLE-CONSTRUCTION FLOW, CFPS 2020
============================================================================
  Full CFPS 2020 adult file                      N = 28,530
  Working age 16-60                              N = 19,707
  Positive labor income (emp_income>0)           N = 9,551
  Drop missing/anomalous valid education         N = 9,034
  Drop missing/anomalous valid health            N = 8,946
  Drop missing/anomalous valid urban/rural       N = 8,309
  Drop missing/anomalous valid marital status    N = 7,845
  Non-missing community & province id            N = 7,845
  Drop singleton communities (LOO IV undefined)  N = 6,948
  (= final regression sample of 6,948; no further observations lost)
