The Role of the Media in Agenda Setting: The Case of Long-Term Care Rebalancing

Abstract

This study investigates the role of print media in state policy agendas in four states—Connecticut, Minnesota, Oregon, and Utah—in rebalancing long-term care away from institutions toward home- and community-based (HCBS) services. Ordinary least squares regression is used to model states’ policy agendas, as measured by the proportion of Medicaid long-term care spending on HCBS expenditures and number of rebalancing bills proposed, from 1999 to 2008. Results reveal a relationship between states’ rebalancing agendas and the extent of media coverage, and state economic, political, and programmatic characteristics. Findings suggest that media coverage reflects broader shifts in state-level attitudes toward rebalancing.

INTRODUCTION

Because nursing homes are a required benefit under the Medicaid program, they have historically received the lion’s share of spending on long-term services and supports (LTSS), rather than alternative settings at home and in the community. However, a variety of factors have favored “rebalancing”—a shift in LTSS provision and spending away from institutions and toward home- and community-based services (HCBS)—but the degree of this shift varies across states. This article examines the extent to which the media operates as a key factor in state-level policy actions to rebalance Medicaid LTSS.

The variation among states in their support of noninstitutional options, collectively known as HCRS, is extreme: spending on HCBS (as a proportion of total LTSS expenses) ranged from 4.4% in Rhode Island and 10.2% in North Dakota to 62.1% and 78.7%, respectively, in Washington and New Mexico (Eiken, Sredl, Burwell, & Gold, 2011). Such variation is possible because states may use several mechanisms under the Medicaid program to make HCBS available: although they are required to provide very limited HCBS under the mandatory home health benefit (which emphasizes skilled nursing services), states have a choice of whether or not to cover adcLitional HCBS that provide a broader range of services. They can do so by choosing to provide personal care as an optional state plan service. They may also take advantage of Medicaid HCBS waiver options, created by Congress in 1981, which allow states to waive Medicaid requirements to offer a broad range of medical, social, and supportive—but not housing—services (Miller, 2002). States often prefer waivers to the personal care option because, by using waivers, they may offer a broader range of services while restricting the populations eligible for them.

Nationally, the shift toward HCBS has been marked and consistent: participants and expenditures grew from 1.9 to 3.3 million and $17 billion to $50 billion, respectively, between 1999 and 2009 (Howard, Ng, & Harrington, 2011; Ng, Harrington, Muscumeci, & Reaves, 2012). Several national-level factors were important in this shift: most importantly, the U.S. Supreme Court’s 1999 Olmstead v. L.C. decision required states to provide supportive services in the “least restrictive environment” possible—that is, in community-based rather than institution settings (Miller, 2005). Over the years, the federal government has increasingly sought to provide states with flexibility in how the Medicaid program can be used to offer HCBS by, for example, integrating participant-directed services into the program. More recently, the Patient Protection and Affordable Care Act (ACA) has given states additional options for expanding access through a combination of enhanced Medicaid matching payments (e.g., State Balancing Incentive Payment Program, Sec. 10202), demonstrations (e.g., Money Follows the Person Rebalancing Demonstration, Sec. 2403), and new Medicaid state plan options (e.g., Community First Choice Option, Sec. 2401; Justice, 2010).

However, these opportunities are meaningful only to the extent that states take advantage of them. Clearly, many factors—economic, political, institutional, and demographic—are likely to influ ence the decisions states make about rebalancing. A considerable body of literature, however, addresses the role of the media in policy making; that is, how and why certain issues rise to the top of the political agenda and become the subject of political action and debate (Kingdon, 199.5; McCombs & Shaw, 1972). It is theorized that the media both shapes political messages and defines problems—their causes and solutions—for the general public and government officials (Baumgartner & Jones, 1993; Dearing & Rogers, 1996; Kingdon, 1995; McCombs, 2004; Weaver, Mccombs, & Shaw, 2004). It does so, in part, through “framing” (Entman, 1993), including: the prominence with which a story is presented; its emotional resonance or affect (e.g., positive, negative, neutral) ; and the specific subtopics addressed (Ghanem, 1997). In general, extant research suggests a direct, reciprocal 'relationship between the media and policy agendas (Dearing & Rogers, 1996).

Just a handful of studies have examined portrayal of the LTSS sector in the mass media (Mebane, 2001; Grogan & Patashnik, 2003; Smith, 1981; Ulsperger, 2002), typically by analyzing print articles in newspapers. One study, for example, found that most coverage (38% of articles) focused on nursing homes; only 13% of the articles were published on the front page (Mebane, 2001). On a related note, Grogan and Patashnik (2003) identified changes in the way Medicaid was characterized in The New York Times, from a residual welfare program for the poor in 1981 to a core social entitlement for low-income families, the aged, and disabled in 1995, during which more emphasis was placed on the program’s coverage of LTSS for ordinary seniors. Given the paucity of existing research on the association between the media and LTSS policy, this study adds considerably to our understanding of how the rise of LTSS as an issue in the media has impacted public policy at the state level. It therefore complements an ongoing body of work by Miller and colleagues, which analyzes coverage of nursing homes in four national print newspapers (Miller, Tyler, Rozanova, & Mor, 2013; Miller, Tyler, and Mor, 2013), but extends this by analyzing the association of media coverage with policy actions in four case study states.

A number of studies demonstrate an association between the media and policy agendas (Baumgartner & Jones, l993; Dearing & Rodgers, 1996; Kingdon, 1995; Tan & Weaver, 2009). For this study, we hypothesized that the proportion of negative articles about nursing homes would be inversely associated with the proportion of Medicaid LTSS dollars spent on HCBS: that is, in states with a relatively greater investment in HCBS we predicted that the press would be less likely to publish negative articles about nursing homes. Moreover, the proportion of negative articles was expected to be directly related to the number of proposed bills addressing rebalancing; that is, states with negative press about nursing homes would be more likely to pursue efforts to rebalance long-term care toward HCBS.

It was further expected that the total number of nursing home-related articles would be directly associated with the proportion of Medicaid LTSS spending devoted to HCBS, since more coverage of nursing homes would reflect awareness of nursing homes issues more generally, and the need to shift spending toward HCBS. It also was expected that the total number of nursing home articles would be directly associated with the number of proposed bills about rebalancing.

Lastly, we expect that the proportion of articles addressing rebalancing issues would be positively associated with the proportion of LTSS spending that goes toward HCBS, since this higher spending on HCBS would likely attract or reflect media attention. For the same reason, a higher proportion of articles addressing rebalancing was expected to be associated with the number of rebalancing bills promulgated.

METHODS

Sample

Case study states—Minnesota, Connecticut, Oregon, and Utah–were chosen to ensure maximum variation in long-term care policy and market characteristics. Using a selection method developed by Miller, Mor, Grabowski, and Gozalo (2009) and data from AARP (Houser, Fox-Grage, & Gibson, 2009), states were ranked according to their Medicaid per diem nursing home reimbursement in 2009 (”high” or “low”) and the percentage of the state’s total Medicaid long-term care budget devoted to HCBS for older people and adults with physical disabilities in 2007 (”high” or “low”). Connecticut had high nursing home reimbursement and low proportion of state Medicaid long-term care spending devoted to HCBS; Oregon had low nursing home reimbursement and high proportion of state Medicaid long-term care spending devoted to HCBS; and Minnesota and Utah were high and low on both parameters, respectively. In all cases, consistent with national trends, Medicaid HCBS spending increased markedly over the time period of the study—nearly doubling, in the case of Connecticut, and increasing by over five times, in the case of Utah.

Data

Articles were selected from a single print newspaper in each state, over a 10-year period (1999 to 2008): all were newspapers of the capital city. Selected dailies included The Star Tribune (Minnesota), The Hartford Courant (Connecticut), The Statesman Journal (Oregon), and The Salt Lake Tribune (Utah). Using LexisNexis, pertinent articles were identified using the following key words: nursing home, nursing homes, nursing facility, nursing facilities, long-term care facility, and long-term care facilities. Articles were excluded if nursing homes were only referenced peripherally (for example, in an obituary). The total number of articles extracted ranged from 127 in Utah to 930 in Connecticut; the number for Oregon (213) and Minnesota (715) fell in between.

Next, the content of LTSS-related media reports in these articles was analyzed. A coding instrument was developed to systematically abstract information, via an inductive process: reading the articles, investigators independently generated suggestions for categories observed in the data. These categories were then further refined over several iterations, in collaboration with other members of the research team (Glaser & Strauss, 1967; Miles & Huberman, 1984), culminating in the final coding instrument. Substantial effort was invested in the coding process, and interrater reliability was assessed and steps taken to ensure agreement; an overall intercoder agreement of 85% was achieved.

The media agenda acted as our key predictor variable: each article was coded for tone and theme. Articles were categorized as having either positive, negative, neutral, or mixed tone. They were also categorized according to theme: whether they addressed such issues as quality, financing, cost, rebalancing, access, legal issues, business/property concerns, and/or natural disasters. Three focal independent variables were used in the present analyses: the total number of nursing home-related articles, the proportion of articles with negative tone, and the proportion of articles focusing on rebalancing themes.

The primary dependent variable, state policy agenda, was measured via two indicators: the proportion of Medicaid long-term care spending directed toward HCBS (Burwell, Sredl, & Eiken, 2009); and the number of proposed bills about rebalancing in state legislatures. As there is no central database for the latter category of data, we derived information about proposed legislation from individual states’ legislative websites (Tan & Weaver, 2009), using the keywords “nursing home(s)” and “home- and community-based services.” Because the distribution of the number of proposed bills was nonnormal, we used the natural log of the number of proposed bills (plus 1 given the presence of zeros) as the dependent variable in our analyses.

Control variables included additional economic, programmatic, and political variables. Three political variables were used: the first is Erikson, Wright, and McIver’s (1993) liberal ideology index. The second is administrative capacity–i.e., the resources available within the public bureaucracy, measured using the number of full time equivalent public welfare employeesper 1,000 population. The third political variable is interest group activity around LTSS, measured using the proportion of the population 65 or older (an indicator of consumer advocacy influence) and the number of nursing home beds per 1,000 elderly population (an indicator of nursing home industry influence; Miller & Wang, 2009). Programmatic variables, which aim to capture structural factors likely to influence the relative availability of institutional and home- and community-based services, include the presence of a Certificate-of-Need (CON) program and/or moratoria, which aim to control the supply of nursing home beds and tend to be associated with lower nursing home spending; and home care availability, measure using the number of home health agencies is per 100,000 population. The state’s unemployment level and per capita gross state product acted as additional economic control variables. In addition, we include a linear time trend to account for year, and state, to control for state-level fixed effects (using Utah as the reference state).

Analytic Plan

We began by examining descriptive statistics on all dependent and focal independent variables. We then calculated Pearson correlations to examine the relationship between each of the four dependent variable and all of the independent variables, checking for multicollinearity. Third, we used ordinary least squares regression to examine the relationship between the two dependent variables and three media variables studied, controlling for potentially confounding factors. Six models were estimated, one for each media variable-dependent variable combination. Due to the small number of observations gathered, the number of independent variables included on the right hand side of each model had to be limited. Six were chosen in addition to the media variable included based on bivariate associations among the variables analyzed . These were unemployment rate, liberal ideology, nursing home beds per 1,000 elderly, proportion of the population 65 years or older, certificate-of-need/moratoria, and home health agencies per 100,000 population.

RESULTS

We began by examining descriptive statistics for each dependent variable (across all years and states), shown in Table 1. Average yearly Medicaid HCBS expenditures were $266 million; the proportion of total Medicaid spending on HCBS ranged from 6.0% to 65.0% annually across the states examined, with an overall mean of 28.6%. The average yearly number of proposed rebalancing bills was 4.8, ranging from 0.00% to 100.0% of all LTSS bills proposed each year across the case study states for an overall average of 13-3%. As expected, the average proportion of Medicaid LTSS devoted to HCBS was highest in Oregon, followed by Minnesota, Connecticut, and then Utah. The percent of proposed LTSS legislation pertaining to rebalancing, on the other hand, was distributed quite differently, with Utah averaging the highest at 37.8% and Oregon the least at 1.2%. In all, there was a total 192 rebalancing bills proposed over the 10-year period studied, ranging from just 15 and 16 bills in Utah and Oregon, respectively, to 74 in Connecticut and 87 in Minnesota (not reported in Table).

TABLE 1.

Mean Medicaid HCBS Expenditures and Proposed Rebalancing Legislation by State, 1999–2008 (n = 40)

	Connecticut	Minnesota	Oregon	Utah	All states
Medicaid HCBS  expenditures ($)	263 million	482 million	283 million	14 million	266 million
Medicaid HCBS  expenditures (%)	19.59	34.06	50.820	10.13	28.64
Proposed rebalancing  bills (no.)	7.40	8.79	1.60	1.50	4.80
Rebalancing bills total  LTC bills (%)	6.89	8.28	1.16	37.76	13.77

Note. HCBS = home- and community-based services; LTC = long-term care.

Table 2 reports descriptive statistics on all independent and control variables, both overall and for each state. Clearly, Connecticut has the most active press with respect to nursing home coverage, whereas Oregon has by far the highest proportion of articles on rebalancing, while Utah has the highest proportion of negative articles about nursing homes. States also vary considerably on many of the control variables: Utah is the most conservative state, as indicated by its score on the ideology index and is also the youngest, while Oregon has the highest unemployment rates and greatest number of public employees. Oregon also has the fewest nursing home beds per 1,000 older people, while Connecticut has the most. Minnesota ranks high on both the nursing home and home health agency supply measures.

TABLE 2.

Independent and Control Variables by State, 1999–2008 (n = 40)

	Connecticut	Minnesota	Oregon	Utah	All states
Total LTC articles	951	714	214	125	2,004
Total negative NH  articles	378	313	76	56	823
Negative NH articles  (%)	40.46	43.46	36.51	51.64	43.02
Total articles on  rebalancing	79	102	78	11	270
Articles on rebalancing  (%)	10.28	14.19	38.95	13.66	19.27
Average  unemployment rate	4.09	4.14	6.26	4.40	4.72
Average per capita  gross state product  ($)	115,646	48,789	41,275	39,070	61,195
Average ideology  index	−0.03	−0.06	−0.16	−0.27	−0.13
Average FTE public  welfare employees  per 1,000	139.55	54.14	164.95	133.38	123.01
Average % 65+  population	13.67	12.17	12.93	8.68	11.86
Average nursing home  beds per 1,000 65+	65.82	64.21	31.87	39.80	50.42
Average home health  agencies per 100,000	2.51	4.62	1.77	2.27	2.79

Note. Where “average” is used, we are averaging annual figures over all years of the study. Where “total” is used, we are totaling figures across all years. NH = nursing home; HCBS = home- and community-based services; LTC = long term care; FTE = full-time equivalent.

Table 3 reports bivariate correlations. Some of our expectations about how the media variables were associated with policy actions were confirmed. Thus, the total number of nursing home articles was directly correlated with the number of proposed rebalancing bills (r = .746, p < .001) whereas the proportion of articles about rebalancing was directly associated with the proportion of Medicaid LTSS spending on rebalancing (r = .499, p < .001).

TABLE 3.

Bivariate Associations Between Dependent and Policy Agenda Variables (n = 40)

	Medicaid expenditures on HCBS (%)	Rebalancing bills (no.)
Media variables
Total nursing home articles	−.09	.746***
Negative articles (%)	−.192	−.05
Rebalancing articles (%)	.499***	−.391
Economic variables
Unemployment rate	.579***	−.412**
Per capita gross state product	−.075	.169
Political variables
Liberal ideology index	.214	.574***
Public employees/1,000 (no.)	.183	−.525***
Population 65+ (%)	.531***	.401**
Nursing home beds/1,000 65+ (no.)	−.282	.624***
Programmatic variables
Certificate-of-need	.414**	.254
Home health agencies/100,000 (no.)	−.07	.592***

Note. HCBS = home- and community-based services.

^**p < 0.1,

^***p < .001.

As expected, the proportion of Medicaid expenditures devoted to HCBS is positively correlated with unemployment rate (r = .579, p < .001), as is the percentage of the population 65 years or older (r = 0.531, p < .001). Also as expected, the proportion of Medicaid HCBS expenditures is positively correlated with the presence of a CON/moratorium (r = .414, p < .05).

Consistent with expectations, the number of rebalancing bills was positively correlated with liberal ideology (r = .574, p < .001), home health agencies/100,000 (r = .592, p < .001), and percent 65+ (r = .401, p < .01). The number of rebalancing bills was also associated with public employees/1,000 (r = −.525, p < .001) and nursing home beds/1,000 65+ (r = .624, p < .001).

Examining correlations allowed us to exclude the following variables from the multivariate analyses: gross state product (because it did not prove significantly related to either dependent variable examined) and public welfare employees/1,000 (because it was highly correlated with the number of home health agencies per 100,000 (r = −.932, p < .001; multicollinearity is defined as Pearson’s r > .700). Thus, in addition to the media variables, the final regression analyses included unemployment rate, liberal ideology index, proportion population 65+, nursing home beds/1,000 65+, certificate-of-need/moratoria, and home health agencies/100,000 population.

Tables 4 and 5 show results for the six regression analyses conducted. Each model explains a high proportion of the variance, with adjusted R-squares in the .600 to .943 range. The media agenda variable proved significantly related in two of the six models.

TABLE 4.

Regression Results: Percent Medicaid Long-Term Care Expenditures Directed Toward HCBS by Media Variables (n = 40)

	Model 1	Model 2	Model 3
Media variables
Rebalancing articles (%)	−.262**	—	—
Negative articles (%)	—	.059	—
Total nursing home articles	—	—	.021
Economic variables
Unemployment rate	.915	1.695#	1.513
Political variables
Liberal ideology index	28.504#	34.545*	41.540*
Nursing home beds/1,000 65+ (no.)	−.052	−.141	−.159
Population 65+ (%.)	10.792*	14.544*	13.084*
Programmatic Variables
Certificate-of-need	8.603*	−.643	3.740
Home health agencies/100,000 (no.)	−4.682	−5.502	−3.682
State (reference is Utah)
Connecticut	−54.363*	−64.662*	−64.022*
Minnesota	−11.995	−16.106	−20.648
Oregon	−11.153	−30.494	−27.617
Linear trend	1.777***	1.366**	1.669**
Constant	−76.945	−106.571	−95.295
Adjusted R2	.943	.926	.925
F-test	59.827***	45.415***	44.612***

^#p < .10,

^*p < .05,

^**p < .01,

^***p < .001.

TABLE 5.

Regression Results: Natural Log of Number of Rebalancing Bills by Media Variables (n = 40)

	Model 1	Model 2	Model 3
Media variables
Rebalancing articles (%)	3.174	—	—
Negative articles (%)	—	.002	—
Total nursing home articles	—	—	.009*
Economic variables
Unemployment rate	2.500	−.006	−.062
Political variables
Liberal ideology index	2.945	−.385	.490
Nursing home beds/1,000 65+ (no.)	−.028*	−.024*	−.024*
Population 65+ (%)	.800	.706	.186
Programmatic variables
Certificate-of-need	−.636	−.606	−.247
Home health agencies/100,000 (no.)	.507	.425	.576
State (reference is Utah)
Connecticut	−1.714	−1.228	.090
Minnesota	−1.572	−1.040	−.566
Oregon	−3.325	−2.615	−.651
Linear trend	.025	.024	−.651
Constant	−6.218	−5.435	−1.221
Adjusted R2	.656	.600	.606
F-test	7.773***	6.328***	6.463***

^*p < .10, p < .05,

^***p < .001.

Contrary to expectations, Table 4 shows that the proportion of Medicaid LTSS spending directed toward HCBS is negatively related to the proportion of nursing home-related newspaper articles about rebalancing (b = −.262, p < .05). The total number of articles and the proportion of nursing home-related newspaper articles with negative tone, however, did not prove to be statistically significant predictors in this area.

Consistent with expectations, there was an association between some of the political, programmatic, and economic variables and the proportion of Medicaid LTSS spending on HCBS. Each model showed that liberal ideology was positively related to the proportion of Medicaid LTSS spending on HCBS (b = 28.504, p < .10; b = 34.545, p < .05; b = 41.540, p < .05). Similarly, population 65+ was positively related to the proportion of Medicaid LTSS spending on HCBS in each model estimated (b = 10.792, p < .05; b = 14.544, p < .05; b = 13.084, p < .05). In addition, one model showed a direct association between the proportion of Medicaid LTSS spending on HCBS and unemployment (b = 1.695, p < .10), while another showed a positive association between the proportion of Medicaid LTSS spending spent on HCBS and efforts to slow the expansion of nursing homes (CON program or moratoria) (b = 8.603, p < .05). There was also a dear state effect and linear time trend in all three models.

Although neither the proportion of articles about rebalancing or with negative tone proved significant, the total number of nursing home articles proved positively related to the natural log of the number of rebalancing bills promulgated (b = .009, p < .05), as expected. Also, in each model, the number of nursing home beds/1,000 persons aged 65+ was negatively related with rebalancing bills (b = −.028, p < .10; b = −.024, p < .10; b = −.028, p < .10).

DISCUSSION

Results suggest that the media agenda may be related to state health policy agendas pertaining to rebalancing LTSS. Conceptually, results reinforce the inferences of political scientists and media scholars who highlight the role of the media in the agenda setting process (Baumgartner & Jones, 1993; Dearing & Rodgers, 1996; Kingdon, 1995; Mccombs, 2004; Tan & Weaver, 2009; Weaver et al., 2004). Empirically, they build on the prior studies characterizing newspaper coverage in the LTSS sector (Mebane, 2001; Grogan & Patashnik, 2003; Miller, Tyler, Rozanova, et al., 2013; Miller, Tyler, & Mor, 2013; Smith, 1981; Ulsperger, 2002).

The media has, until now, rarely been studied within the context of more commonly identified predictors of state health policy agendas, which include various economic, political, and programmatic variables (Miller, 2004, 2005; Berry & Berry, 2007). This study adds to that literature by looking at the relationship between media coverage and state health policy making—specifically, efforts to rebalance Medicaid LTSS away from expensive and less preferred institutional services and toward HCBS—while controlling for other predictors of state health policy agendas. It found that in two models, the media was a statistically significant correlate of states’ rebalancing agendas—a substantial and promising finding, particularly in light of the study’s exploratory nature and limited sample size. Each of the models estimated accounted for a high proportion of variance, with adjusted R-squares indicating that 60% to 94% of state agendas pertaining to rebalancing could he explained by the variables examined.

The three focal media variables measured different aspects of press coverage of LTSS issues. The proportion of articles focusing on rebalancing issues was inversely associated with the proportion of Medicaid LTSS long-term care spending directed toward HCBS. In addition, the total number of nursing home-related articles was found to be positively associated with the natural log of the number of rebalancing bills in a state. Together these results suggest a relationship between print media coverage and the policy agenda, as identified in previous studies (Baumgartner & Jones, 1993; Beckett, 1997). However, the former relationship did not play out in the manner hypothesized in this study: we expected that the proportion of articles addressing rebalancing issues would be positively associated with the proportion of LTSS spending going toward HCBS, assuming that a higher proportion of HCBS spending would likely attract or reflect media attention. Instead, we found that media attention on rebalancing was higher (lower) when the proportion of LTSS spending on rebalancing was lower (higher), possibly because the action and debate surrounding a policy deficit or need—in this, case, insufficient HCBS spending relative to institutional spending—may be more likely to generate press coverage than when that deficit or need has been satisfied.

The study did not confirm the media’s influence in the context of some of the other relationships examined. For example, the number of nursing home-related articles was significantly associated with the number of rebalancing bills but not the proportion of Medicaid LTSS spending on HCBS. This suggests that the total amount of media coverage may be more likely to influence or reflect policy processes (rebalancing bills proposed) than outcomes (state Medicaid spending on HCBS relative to nursing homes). That the proportion of articles about rebalancing hills proved significantly associated with the proportion of HCBS spending but not proposed rebalancing bills suggests the opposite conclusion where the content of media coverage is concerned. Current Medicaid spending reflects previously made policy decisions, whereas proposed legislation reflects policy makers’ current responses to the LTSS policy environment. The association between states’ rebalancing agendas and the rebalancing content of the media coverage reported is more robust when the policy environment measure reflects past legislative decisions (spending) as opposed to current legislative deliberations (proposed bills). In this case, at least, it appears that the total amount of media coverage is more reflective of legislative proposals than outcomes whereas the specific content of media coverage is more responsive to legislative outcomes than proposals.

In addition to the media, this study modeled state policy relating to rebalancing and several other state characteristics—including various economic variables (unemployment), political variables (liberal ideology, nursing home industry power, consumer advocacy power), and programmatic variables (nursing home supply restrictions, home care availability). Where significant, signs on each of these characteristics fell in the directions one might expect.

In one model, the unemployment rate was found to be positively associated with an increase in the proportion of Medicaid LTSS spending devoted to HCBS. This finding suggests that declining fiscal health may lead a higher proportion of the population to seek out public assistance (i.e., Medicaid HCBS funding). It might also suggest that states in fiscal distress might seek ways to reduce spending relative to current levels, by, perhaps, substituting less expensive HCBS for more expensive institutional care. The relationship between unemployment and state policy making with respect to long-term care rebalancing confirms findings observed in several previous studies focusing on state policy in the health sector more generally (Berry & Berry, 2007; Miller & Wang, 2009, Miller, 2005).

In three models liberal ideology was a strong, positive predictor of the proportion of Medicaid funding allocated to HCBS. This result indicates that states with more liberal policy makers and voters were more likely to champion public assistance programs, not to mention the redistribution of wealth necessary for doing so. It also suggests that shifting Medicaid long-term care spending away from nursing homes and toward home- and community-based settings, a break in past policy practices, may be more likely in such states. Such an association has been found previously (Miller, 2005; Miller & Wang, 2009; Erikson, Wright, & McIver, 1993). The same three models also reveal strong positive correlations between the proportion of the population 65 years or older, proxy for elder advocacy power, and the proportion of Medicaid LTSS spending directed toward HCBS, This result indicate5 that older states were more likely to pursue long-term care rebalancing than states with younger populations. It is consistent with surveys revealing strong preferences among elders to remain in the least restrictive setting possible, as long as possible, typically at home rather than in a nursing home, should the need for LTSS arise (AARP, 2003; “Los Angeles Times poll,” 2009). It is also consistent with prior research revealing an association between interest group activity on behalf of the elderly and long-term care policy outcomes at the state level (Harrington, Mullan, & Carrillo, 2004; Miller, 2006).

The number of nursing home beds per 1,000 population 65 years or older served as a proxy for nursing home industry power. In three models we observed a negative relationship between the number of heels and the number of rebalancing bills proposed; suggesting that greater nursing home industry power would be associated with a lower commitment to HCBS. The relationship between the number of nursing home beds and state progress in rebalancing confirms earlier findings (Grogan, 1999; Miller & Wang, 2009, Miller, 2005).

Limitations

The study has some limitations. First, the search terms used to identify newspaper articles focused on nursing homes and may have excluded relevant articles focusing on rebalancing. However, it is reasonable to assume that most, if not all articles that mention rebalancing also mention nursing homes. Second, although we studied four very different states over an extended time period, results may have differed if a different combination of states and time period had been chosen. Third, the small sample size required us to exclude some potentially important control variables. And fourth, we could only examine association, not causation, given the pooled cross-sectional nature of the analyses. We sought to explore causality, in part, by examining the dependent variables in one year as a function of the key media variables during the previous year. Lagging the media variables did not prove fruitful, however; doing so resulted in even smaller sample sizes than those ultimately employed, making it even more difficult to detect statistically significantly associations in the data.

CONCLUSION

In sum, the present study confirmed expectations that the media deserves the same level of attention given to economic, programmatic, and political antecedents of state policy agendas. In order to more thoroughly investigate this relationship in relation to rebalancing, however, it is necessary to supplement the investigation reported here. This would include incorporating data from a larger numbers of states over a longer period of time. It would also be beneficial to incorporate interviews with key stakeholders, to provide deeper insight than is possible using quantitative data alone. Interviews could be used for purposes of triangulation, reinforcing, or qualifying the statistical results reported, in addition to better understanding the mechanisms that underlie the associations identified between the media and state policy agendas in the current study. Last, it is important that the relationship between the media and state policy agendas be explored in different, unrelated contexts. This can be in relation to long-term care, the health sector more generally, and other policy areas altogether.