Table 3.

Instrumental Variables Estimates of the Relationship between Methamphetamine Consumption and Drug Treatment Admissions

	Log methamphetamine treatment (1)	Log cocaine treatment (2)	Log heroin treatment (3)	Log cannabis treatment (4)	Log alcohol treatment (5)	Log other treatment (6)
Log amphetamine hospitalizations	0.292 [0.098]	−0.013 [0.070]	−0.084 [0.100]	−0.157 [0.232]	−0.283 [0.158]	−0.284 [0.178]
Log cocaine hospitalizations	−0.026 [0.030]	0.062 [0.032]	−0.004 [0.022]	0.02 [0.051]	0.083 [0.038]	−0.004 [0.049]
Log opioid hospitalizations	−0.007 [0.023]	−0.026 [0.033]	0.062 [0.028]	0.029 [0.038]	0.046 [0.036]	0.047 [0.030]
Log cannabis hospitalizations	−0.007 [0.037]	0.059 [0.039]	0.001 [0.026]	0.044 [0.063]	0.104 [0.038]	0.167 [0.056]
Log alcohol hospitalizations	−0.136 [0.044]	0.014 [0.041]	−0.029 [0.040]	−0.015 [0.058]	0.023 [0.043]	0.048 [0.062]
Constant	87.537 [57.579]	100.762 [55.893]	23.029 [53.705]	94.985 [66.673]	45.104 [62.526]	56.106 [55.223]
Year and month dummies	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2,784	2,784	2,784	2,784	2,784	2,784
R2	0.95	0.98	0.99	0.92	0.97	0.94

Notes: All of the regressions include county fixed effects, and control for the proporation black and Hispanic, and for the proportion age 15–19, 20–24, 25–29, 30–34, and 35–39. Standard errors are in brackets; they are robust and clustered on county. The proxy for consumption is the log of the count of amphetamine-related hospital admissions in a county in a month + 1. The unit of observation is county by month. The instrument takes on a value of one between August 1995 and September 1996 and the estimates are robust to varying the starting and ending date.