Four phases of the Flint Water Crisis: Evidence from blood lead levels in children

Abstract

The Flint Water Crisis (FWC) is divisible into four phases of child water-lead exposure risk: Phase A) before the switch in water source to the Flint River (our baseline); Phase B) after the switch in water source, but before boil water advisories; Phase C) after boil water advisories, but before the switch back to the baseline water source of the Detroit Water and Sewerage Department (DWSD); and Phase D) after the switch back to DWSD. The objective of this work is to estimate water-lead attributable movements in child blood lead levels (BLLs) that correspond with the four phases in the FWC. With over 21,000 geo-referenced and time-stamped blood lead samples from children in Genesee County drawn from January 01, 2013 to July 19, 2016, we develop a series of quasi-experimental models to identify the causal effect of water-lead exposure on child BLLs in Flint. We find that the switch in water source (transitioning from phase A to B) caused mean BLLs to increase by about 0.5μg/dL, and increased the likelihood of a child presenting with a BLL ≥ 5μg/dL by a factor of 1.91-3.50, implying an additional 561 children exceeding 5μg/dL. We conservatively estimate cohort social costs (through lost earnings alone) of this increase in water-lead exposed children at $65 million, contrasted with expected annual savings of $2 million from switching water source. On the switch from Phase B to C, we find BLLs decreased about 50% from their initial rise following boil water advisories and subsequent water avoidance behaviors by households. Finally, the return to the baseline source water (Phase D) returned child BLLs to pre-FWC levels further implicating water-lead exposure as a causal source of child BLLs throughout the FWC.

Keywords: Blood lead levels; Child BLLs; Child health; Flint Water Crisis; Lead exposure.

Figure 1

Lead exposure phases of the flint water crisis.

Figure 2

Google trends search interest in Flint-Saginaw-Bay City region of Michigan. Search for “Notes: Flint water” presented at a daily time-step; “Flint water contamination” and “water boil” presented at weekly time-step. Flint water (daily time-step); Flint water contamination and water boil (weekly time-step)

Figure 3. The City of Flint and Genesee County study areas

Figure 4

Difference-in-differences coefficient vis-a-vis child blood lead (μg/dL) by elapsed time since the start of Phase B (May 25, 2014, t = 0). Coefficients derived from model reported in Table 2, Column 1, involving the incremental expansion of the post-period.

Figure 5

Predicted blood lead levels (μg/dL) and probabilities of ≥5 μg/dL before the switch to Flint River water (A) and after the switch back to Detroit water (B). Predicted values in Panel A from Table 7, Model and from Table 7, Model 6 corresponding to the sample restricted switch back period (before the surge in blood lead testing). Predicted values in Panel B from Table 8, Model 1 and from Table 8, Model 6 corresponding to the sample restricted switch back period (before the surge in blood lead testing). All other covariates are fixed at sample means. F= Flint, NF = Not Flint, FR = Flint, Sample Restricted, NFR = Not Flint, Sample Restricted

Figure 6

Predicted child blood lead levels (μg/dL) by month, Michigan-wide, 1999-2012. Predicted values derived from a LS model of 1,180,928 children residing in one of eighty-three counties in Michigan from 1999 to 2012. In addition to fixed effects for month, model covariates in included child age, child sex, as well as county and year fixed effects. Confidence intervals by delta-method standard errors.