Washington State’s public health system is similar to that of other states. It is de-centralized, which means LHDs are primarily led by employees of local governments; each LHD is, in turn, overseen by a local board of health, which selects and oversees the professional leadership of the local public health agency, adopts local ordinances and resolutions, and approves budgets.^14,15 LHDs in Washington State are funded largely through local state, federal government sources; in 2014, local government contributions accounted for approximately 15% of revenues; local licenses and fees, 30%; federal grants (direct and indirect), 28%; state grants and funding sources, 22%; and other sources, 5% (data not shown)

The principal data used in this study come from the Public Health Activities and Services Tracking (PHAST) project (see http://phastdata.org/about). These data, which represent cleaned and compiled annual financial data from Washington State’s Budget, Accounting, and Reporting System, have been used by the PHAST team in various studies.^16–19 The portion of the PHAST data set used here included revenues and expenditures reported annually between 1998 and 2014 by LHDs operating in Washington State (35 LHDs). Additional county-level and LHD structural-level variables were taken from the Area Health Resource Files, Bureau of Labor Statistics Current Population Survey, American Community Survey, and Census Bureau Small Area Health Insurance and Income/Poverty Estimates.

In 2007, the Washington State Legislature passed Engrossed Second Substitute House Bill (E2SSHB) 5930. The law included $20,000,000 in new state appropriations—available in $10,000,000 tranches for each of the state’s fiscal years 2008 and 2009—for LHDs. LHDs, in turn, were required to use those new state 5930 dollars to support “core public health functions of statewide significance,” as defined by an expert panel with representation from LHDs, community service providers, and other stakeholder groups.^20,21

The 5930 program presents a unique opportunity to examine the effects of new state funding on LHD expenditures. First, there is a clear onset date for the funding, and it is both preceded and followed by multiple years of data for each LHD. Second, the state operates on a biennial budget, so the amount of funding an LHD would receive was typically determined in the fiscal year prior to its availability and then fixed for the next two fiscal years. An LHD would know in fiscal year 2007 how much 5930 funding it would receive from the state in fiscal years 2008 and 2009. Third, 5930 dollars were distributed according to a statutorily determined formula, and elements in that formula may be used as instruments in sensitivity analyses. Fourth, the authorizing legislation specified that recipient jurisdictions “shall not supplant existing local funding with such state-incented resources.” ²² In essence, this means that the state 5930 program includes a built-in structural bias against the kinds of substitution effects predicted by standard economic theory. Finally, funding was maintained (albeit at reduced levels) as a separate line item through 2013, so it is possible to examine both short- and long-term responses in LHD spending.

Accordingly, the primary independent variable was the amount of per capita revenue that an LHD received through the state 5930 funding mechanism in a given year.

Previous studies of inter-governmental grants and their effects on recipient expenditures have typically defined the outcome of interest in terms of either per capita “own source” revenues or spending.²³ To ensure comparability between the results of this research and previous reported findings, the outcome of interest in this analysis was defined as total per capita expenditures by an LHD in a given year.

The analytic model used in this study controls for a parsimonious set of community characteristics and LHD financial circumstances that theory or previous research indicates may affect LHD access to funding and responses to changes in funding.^24–28 Supplemental Digital Content 1 provides additional details about the definitions and sources of all control variables included in this analysis.

Following work by Leduc and Wilson, who examined state government spending responses to a large increase in federal highway grants under the 2009 American Recovery and Reinvestment Act, this study used a pre-post design that controlled for LHD-specific fixed effects.²⁹ All dollar variables were inflated to constant 2016 dollars using the Consumer Price Index for Medical Care. The baseline year, which remained fixed across ordinary least squares (OLS) regression models, was 2006, as it represented the last full year before the Washington State legislature enacted Engrossed Second Substitute House Bill (E2SHB) 5930.

To examine the immediate impact of the 5930 funding on LHD expenditures, change in per capita LHD spending between 2006 and 2008 was regressed on the corresponding changes in 5930 funding and a set of pre-specified control variables over the same period (Supplemental Digital Content 2 provides the detailed econometric specification). The same regression was repeated to compare changes between 2006 and 2009. Examining changes in per capita LHD spending separately for each of the first two years that 5930 funding was available allowed for the possibility of temporal lags in the local response to newly available state revenues.

To evaluate whether these initial responses in LHD expenditures persisted over time, the change in per capita LHD expenditures between 2006 and 2013 was regressed on the corresponding changes in 5930 funding and the same set of controls. In state fiscal year 2014, the 5930 program ceased to exist as a stand-alone source of state funding, so 2013 represented the last year in which 5930 funds were separately captured in the PHAST data.

The above-outlined analyses assumed that state 5930 allocations to LHDs were not correlated with changes in LHD spending or revenues from local government sources over the same time period. The 5930 funding was primarily allocated according to a statutorily defined formula, which included a per capita amount that was calculated based on historical jurisdiction census information. A strict exogeneity assumption--meaning that the 5930 allocations were independent of spending or revenues from government sources-- was therefore reasonable. Nonetheless, a small proportion of the funding was distributed according to base thresholds designed to ensure that no LHD received less than $100,000. It is possible that those base level funding amounts were determined according to political processes that introduced endogeneity into the relationship between 5930 funding and local responses to it: for example, state legislators representing smaller jurisdictions might only have agreed to support the 5930 funds if their jurisdictions received a base amount that was substantially larger than would have been available through a straight per capita allocation formula.³⁰ To address this possibility, baseline analyses for all three time comparisons of interest were rerun using two-stage least squares (2SLS) regression models employing the use of an instrumental variable (IV). The IV was derived from the aforementioned 5930 funding formula: specifically, the proportion of the total state population residing in an LHD’s jurisdiction in the year used to determine 5930 appropriations (2006 for 5930 allocations in 2008 and 2009 period; 2010 for 5930 allocations in 2012 and 2013). Local per capita LHD expenditures are unlikely to be shaped by a jurisdiction’s relative share of the state population—and this is particularly true when the measure of state population residing in the jurisdiction is collected two to three years prior to funding allocations. The data collection lag thus further supports the case for exogeneity between the IV and local LHD spending decisions. Separate OLS regression models were also run for the “base” and “formula” funding recipients in each of the three time periods of interest.

All data analyses were conducted in Stata version 14 (StataCorp LP, College Station, TX), with significance at P < .05.

Table 1 presents pooled descriptive statistics for LHDs in Washington State in 2006 and 2013, as well as the average change over that time period. Adjusted to 2016 U.S. dollars, average LHD expenditures per capita contrasted sharply between 2006 and 2013 ($72.89 vs. $48.97). However, considerable heterogeneity persisted across LHDs in both years, with average per capita expenditures ranging from $26.63 to $204.67 in 2006, and from $16.59 to $120.29 in 2013.

Sequential cuts to 5930 program funding in the 2010/2011 and 2012/2013 Washington State budgets reduced mean per capita 5930 revenues from $6.10 (S.D. $10.97) for LHDs in 2008 to $3.35 (S.D. $5.89) in 2013 (data not shown). Nonetheless, 5930 funding represented one of the few revenue sources available to LHDs for which the net per capita change over the study period was uniformly positive. Revenues from other state sources, such as Washington State Department of Health (DOH) funding, contracted by approximately $4.00-$5.00 per capita on average.

Early diagnostic analyses indicated that Wahkiakum County experienced uniquely extreme decreases in per capita LHD expenditures over the study period. To prevent bias in the results, Wahkiakum County was excluded from all analyses.

Table 2 presents OLS regression results from each of the three time periods of interest. The coefficients associated with 5930 funding in the first two years of the program were statistically significant (2008: 0.694 [SE: 0.04]; 2009: 0.689 [SE: 0.07]), and indicated that, foreach additional dollar in 5930 funding that LHDs received in 2008 and 2009, net per capita spending increase by approximately $0.69. By 2013, the coefficient associated with 5930 funding was smaller and no longer statistically significant (2013: 0.162 [SE: 0.46]).

Early testing indicated that the proportion of the total state population residing in an LHD’s jurisdiction in the year used to determine 5930 (i.e. the proposed instrumental variable) was a poor instrument when revenue and expenditures were expressed in per capita terms. However, those same tests indicated a strong relationship between the instrumental variable (IV) and log-transformed versions of the independent and dependent variables of interest. To ensure comparability between the base analysis and IV analysis results, the former was rerun using log-transformed inputs and outcomes.

Results from endogeneity tests were consistently statistically insignificant, which suggested that 5930 funding was, in fact, exogenous.

Table 3 presents OLS results separately for LHDs that received the guaranteed base funding amount (“base”) and those that had sufficient population size to qualify for application of the 5930 funding formula. Despite small sample sizes, some differences did emerge—particularly in the second full year of the program. In 2009, each additional dollar of 5930 funding received by formula funding recipients (larger LHDs) was associated with a statistically significant increase of $1.23 in total per capita expenditures. By contrast, among base funding recipients (smaller LHDs), each additional dollar of 5930 funding increased total per capita expenditures by $0.78.

By 2013, point estimates for the effects of 5930 funding had decreased in both groups: each additional dollar of 5930 funding increased total per capita expenditures by $0.951 among formula-funded LHDs and $0.35 among base-funded LHDs (smaller LHDs). However, the standard errors associated with these estimates were large –or, in the case of the based-funded LHDs, inestimable—so neither result was statistically significant.

This study has limitations to note. First, the small sample size limited statistical power in the main analyses, as well as opportunities for robust comparisons across potentially meaningful LHD sub-populations. Despite these limitations, some effects proved sizeable enough to attain statistical significance, and others were qualitatively dissimilar enough to suggest potentially rich areas for future research. Second, this study focused on a single funding program introduced in one state, and that state is somewhat unique in terms of the long-term, county-level, comparable financial data available to researchers and practitioners through PHAST Nonetheless, Washington is similar to many other states with respect to how its local public health system is organized and financed, and the 5930 program was specifically directed towards augmenting their LHDs’ performance of core public health functions.^15,18 The results are thus likely to prove informative to many federal, state, and local policymakers—particularly those interested in identifying financing mechanisms that better support local public health infrastructure. To the extent that other state and local health departments would struggle to replicate these analyses, this research also underscores both the value of, and need for, a uniform chart of accounts for public health departments at all levels—one that delineates “boundaries, linkages and financial flows between state, local, and federal programs.” ^1,34 Finally, like overall LHD expenditures per capita, 5930 funding decreased over the study period, such that per LHD funding available in 2013 was ~60% of that received in 2008 and 2009. Future research will be needed to better understand whether LHD spending responds differently over time when newly introduced sources of funds remain relatively stable.