Table 7.

Heterogeneous impacts in the subscriber sample: decider has no regular pay.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Subsidized Desludging	Unsubsidized Mechanized Desludging	Mechanized Desludging	Any Deposits	Value of Deposits	Any Non-Final Deposits	Value of Non-Final Deposits
Save at will	−0.005 (0.055)	−0.103∗ (0.058)	−0.119∗ (0.066)	−0.049 (0.060)	−0.888 (1.420)	−0.007 (0.038)	0.457 (0.489)
Monthly billing	0.029 (0.053)	−0.129∗∗ (0.056)	−0.088 (0.063)	0.016 (0.058)	0.437 (1.374)
No regular pay	−0.051 (0.046)	−0.038 (0.048)	−0.110∗∗ (0.055)	−0.103∗∗ (0.050)	−1.696 (1.186)	−0.024 (0.032)	−0.068 (0.417)
Save at will × No regular pay	0.074 (0.064)	0.036 (0.067)	0.141∗ (0.076)	0.136∗ (0.070)	2.358 (1.658)	0.048 (0.045)	0.040 (0.578)
Monthly billing × No regular pay	−0.055 (0.063)	0.141∗∗ (0.066)	0.077 (0.075)	−0.014 (0.069)	−0.565 (1.632)
R2	0.325	0.407	0.470	0.315	0.293	0.368	0.304
Mean of Dependent Variable	0.188	0.212	0.406	0.233	4.723	0.067	0.652
N	1479	1365	1365	1479	1479	998	998
Combined Effects:
At will if no regular pay	0.069∗∗ (0.032)	−0.067∗∗ (0.034)	0.022 (0.039)	0.087∗∗ (0.035)	1.471∗ (0.840)	0.041∗ (0.024)	0.498 (0.313)
Monthly if no regular pay	−0.026 (0.034)	0.012 (0.036)	−0.012 (0.040)	0.002 (0.037)	−0.128 (0.878)

Note: Controls in all regressions include: high subsidy and deposit interventions, pre-intervention outcomes (mechanized desludging in year before baseline and manual desludging in year before baseline), unbalanced controls (hhd wealth and hhd size), and fixed effects at the grid-point level. Standard errors are in parentheses: ${}^{\ast }p<$0.10, ${}^{\ast \ast }p<$0.05, ${}^{\ast \ast \ast }p<$0.01. Having regular employment is measured in the baseline. Outcome variables are (1) purchased the subsidized mechanized desludging through the subscription, (2) purchased an unsubsidized mechanized desludging between the baseline and endline, (3) purchased any mechanized desludging between the baseline and endline, (4) used the desludging account, (5) total value deposited in the desludging account (in thousands of CFA, 1000 CFA is approximately $2), (6) number of non-final deposits, and (7) value of non-final deposits (in thousands of CFA). In columns (1) through (5) the excluded treatment group control variable is pay in full. Columns (6) and (7) drop the pay-in-full group, and the excluded treatment group control variable is monthly billing. In the lower panel, ‘At will if no regular pay’ shows the value of the sum of the coefficient on ‘Save at will’ and ‘Save at will x No regular pay,’ the heterogeneous treatment effect.