key: cord-0990099-bsa19i0k
authors: Sapat, Alka; Lofaro, Ryan; Trautman, Ben
title: Policy responsiveness and institutions in a federal system: Analyzing variations in state-level data transparency and equity issues during the COVID-19 pandemic
date: 2022-05-26
journal: Int J Disaster Risk Reduct
DOI: 10.1016/j.ijdrr.2022.103066
sha: ad30e1c221b5faf8882b31f676af439b78034487
doc_id: 990099
cord_uid: bsa19i0k

In the absence of a coherent federal response to COVID-19 in the United States, state governments played a significant role with varying policy responses, including in data collection and reporting. However, while accurate data collection and disaggregation is critically important since it is the basis for mitigation policy measures and to combat health disparities, it has received little scholarly attention. To address this gap, this study employs agency theory to focus on state-level determinants of data transparency practices by examining factors affecting variations in state data collection, reporting, and disaggregation of both overall metrics and race/ethnicity data. Using ordered logistic regression analyses, we find that legislatures, rather than governors, are important institutional actors and that a conservative ideology signal and socio-economic factors help predict data reporting and transparency practices. These results suggest that there is a critical need for standardized data collection protocols, the collection of comprehensive race and ethnicity data, and analyses examining data transparency and reductions in information asymmetries as a pandemic response tool—both in the United States and globally.

antibody tests with PCR testing, meant that the data were not reliable (Kettl 2020 ). Yet, these data were used by states to justify re-openings and to loosen restrictions (Madrigal and Meyer 2020) . Second, to combat health disparities and effectively fight COVID-19, it is critical to understand its impact and spread among communities of color and marginalized populations, given its disproportionate impacts in these communities. But the lack of standardization in data collection and the variations among states in reporting race and ethnicity data made it difficult to target testing and offer protections for those who needed it, particularly in minority communities.

We address these issues by looking at data transparency and equity as seen by state variations in the comprehensiveness of their reporting of health metrics and variations with respect to data disaggregation of race and ethnicity metrics. We view this through the theoretical lens of agency theory, focusing on state political institutions and policy responsiveness to the ideological preferences of citizens, citizen needs, and socio-economic vulnerabilities. Our argument holds that the fragmented nature of COVID-19 policy responses has created differential state-level information asymmetries between public officials (the agents) and the citizenry (the principal). We analyze determinants of statelevel data transparency practices to understand the factors that have contributed to such asymmetries.

A discussion of transparency and agency theory follows, succeeded by a discussion of our data, variables, methods, and findings. Our conclusions suggest policy recommendations to address information asymmetries and promote better data equity and transparency practices. The importance of standardizing data and death counts on a global level and examining the issue in other federalist countries is also briefly discussed.

Governmental transparency is commonly defined as the action of offering citizens access to information through various mechanisms to allow individuals to understand the inner workings of the state, thus enhancing the accountability of public officials to citizens (Piotrowski and Van Ryzin 2007;  J o u r n a l P r e -p r o o f Tejedo-Romero and De Araujo 2018) . Fung (2013) discussed four principles of transparency: information about the actions and operations of the public organizations that affect citizens should be rich and made publicly available; the amount of information made available should be proportionate to the degree to which citizens' interests are jeopardized by public organizations; information should be provided to the individuals and groups most likely to use said information; and political, social, and economic structures should be organized so as to allow citizens to use information to take action. Transparency has the potential to enhance citizen satisfaction, participation, trust in government, and governmental legitimacy, while reducing secrecy and corruption (Cucciniello et al. 2017) . Democratic accountability benefits alongside an increase in transparency, as the dissemination of information gives citizens the opportunity to scrutinize public decisions and decision-making processes (Birkinshaw 2006) . Information, however, is an imperfect public good marred by asymmetries and substantial barriers to access (Stiglitz 2000) , and much like other public sector processes, transparency is subject to faulty and ineffective applications (Cucciniello et al. 2017) . Hence, it is important to analyze the factors contributing to governmental transparency to better understand why asymmetries exist and how they may be reduced.

This study examines determinants of transparency-defined as the accurate collection, disaggregation, and dissemination of COVID-19 data-through the theoretical lens of agency theory.

According to agency theory, a relationship exists between a principal and an agent wherein the agent carries out tasks for the principal; however, the agent has the incentive to hide information from the principal, leading to a loss of welfare (Kiewiet and McCubbins 1991) . In public administration, public officials act as agents on behalf of the citizenry, the principal (Lane 2005) . Problems arise when the relationship between the citizenry and the state encounters information asymmetries that necessitate the use of different mechanisms aimed at holding agents accountable to the principal (Lane 2005).

Releasing information that allows for state actors to be monitored by those they aim to serve increases J o u r n a l P r e -p r o o f the chances of achieving accountability (Laswad et al. 2005) . The result is a reduction in information asymmetries through greater transparency and an enhancement in trust in government (Tejedo-Romero and De Araujo 2018) , in addition to a positive impact on democratic practices such as citizen involvement (Fung, 2013) . Governmental transparency thus possesses the ability to reduce the information asymmetries that hinder democratic practices in the relationship between public officials as agents and the citizenry as the principal.

Transparency can also enhance the legitimacy of governmental actions (see Matheus and Janssen 2020; Tejedo-Romero and De Araujo 2018) . Legitimacy is defined as "a generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions" (Suchman, 1995, p. 574) . Organizations commonly disclose pertinent information to stakeholders to promote legitimacy (Deegan, 2002) . In the public sphere, this increased disclosure and dissemination of information-i.e., the enhancement of transparency and reduction in information asymmetries-has a positive effect on public perceptions of state decisions (Heald, 2006) . The legitimacy of governmental decisions improves alongside transparency mechanisms that give citizens a look into public sector decision-making processes (De Fine Licht et al., 2014) . Information diffusion and dissemination holds public officials accountable to citizens, increases citizen's understanding of public decision-making, promotes an image of good governance, and helps public officials legitimize actions and gain the confidence of the citizenry (De Araujo and Tejedo-Romero 2016). Transparency can be improved by opening government data up to the public, an action which gives citizens a window through which to examine the inner workings of the state and reduces extant information asymmetries, though it should be noted that inaccurate and/or manipulated data may further perpetuate such asymmetries (Matheus and Janssen 2020) . It is thus beneficial for public officials as agents to enhance legitimacy and reduce information asymmetries by accurately collecting and disseminating information to the principal in a transparent fashion.

J o u r n a l P r e -p r o o f

Though the term federalism does not appear in the U.S. constitution, it is one of the most important features of American democracy. While the appropriate roles for the states and the federal government have been fiercely debated, states have often served as "laboratories of democracy" (Kettl 2020, p. 595; quoting former Supreme Court Justice Louis Brandeis). COVID-19 tested the great American experiment of federalism (Benton 2020) by acting as an exogenous shock to the system that has illuminated and amplified various parts of the social world in need of redress (Alexander & Smith 2020) . Although states have acted as the laboratories for this natural experiment, there have been wide variations in how they have managed the pandemic, with more experimentation and very little learning, leading Kettl (2021) to declare that federalism as practiced in the United States has been dangerous to our health.

According to Benton (2020) and Birkland et al. (2021) , intergovernmental relations (IGR) during the pandemic do not reflect a specific form of federalism but can rather be described as a "kaleidoscope" federalism under which no particular strand prevails. As stated by Benton (2018) , "federalism and IGR are no longer either 'cooperative,' 'creative,' 'competitive,' 'coercive,' or 'calculative.' All of these descriptions and approaches, as well as many others, depict the operation and relations of the federal system, even within the same policy area" (19) . IGR are kaleidoscopic in that they "can be identified by metaphors such as fragmented, push-back, nuanced, fend-for-yourself, and collaborative" (Benton 2020, 537) . Inter-local and -state agreements have replaced large-scale collaborations in the absence of clear directions from the federal government (Birkland et al. 2021) .

Political partisanship and polarization have led to fractured forms of federalism characterized by patchwork policies across the country (Bowling and Pickerill 2013) , leading ideology and partisanship to drive regional decisions (Conlan and Posner 2016) . During the pandemic, fragmentation, coupled with the lack of a coherent federal response, led to significant policy variations at the state level (Birkland et al. 2021 ).

State-Level Variations in Data Reporting and Transparency: A coordinated federal response that is critical during crises was missing in the U.S. As noted by Kapucu and Moynihan (2021) , "the lack of national leadership and lack of coordinated action at the federal level caused similar inconsistency among emergency management agencies and health counterparts at state and local levels" (15). Given the polycentric nature of an American crisis response system that calls for collaborative leadership, the absence of a consistent federal strategy (Xu & Basu 2020) created a system where uniformity in data was lacking. At the federal level, a number of pandemic response programs had been shut down or lacked capacity. For instance, the National Security Council (NSC) Directorate of Global Health Security and Biodefense created by the Obama administration in 2016 was disbanded in 2018 under the Trump Administration, and NSC members with expertise in global health security and emergency preparedness were reassigned or left. With respect to data collection, the faulty testing kit the CDC initially developed in February 2020 for states to conduct their own testing and get test results faster than shipping samples to the CDC led to critical delays in identifying COVID-19 cases (Temple-Raston 2021).

With increasing cases and deaths through March and April 2020, there were some attempts at the federal level to improve COVID-19 data collection. For instance, in June 2020, the Department of Health and Human Services (HHS) released data reporting guidance that required the reporting of demographic factors by August 2020 (U.S. Department of Health and Human Services 2020). However, the process lacked adequate guidance or enforcement and only applied to laboratories while not requiring state or local public health departments to report COVID-19 mortality data by any specific demographic breakdowns (Servick 2020; Krieger, Testa, Hanage, and Chen 2020; Zylla and Bernard, 2020) . House Democrats also attempted to introduce legislation on data equity, such as the Equitable Data Collection and Disclosure on COVID-19 Act bill in April 2020 (GovTrack.us 2020a), with a similar bill in the Senate (GovTrack.us 2020b), focusing on the collection and disaggregation of racial and ethnic demographic data for testing, treatment, hospitalization, and fatality results, and the creation of a J o u r n a l P r e -p r o o f Commission on Ensuring Health Equity (H.R. 6585 2020). Despite gaining support from groups such as the AMA (American Medical Association 2020a), both bills died in committee. Other attempts by congressional Democrats such as the COVID-19 Racial and Ethnic Disparities Task Force Act of 2020 bill introduced on May 5, 2020, and the Ending Health Disparities During COVID-19 Act of 2020 introduced on September 11, 2020, also did not get through committees. Data collection on key COVID-19 statistics was thus left mostly to state governments; however, states have varied widely in their efforts.

With regard to the COVID-19 testing strategies that became critical to policy decisions, such as lockdowns, there were various state testing and reporting strategies. In some states, results were reported from public and private labs, while in others, there was great reluctance to report any test results at all (Kettl 2020). States used different definitions for the data they collected, and there were inconsistencies even with their own definitions of metrics. For instance, some states defined their total tests as being the total number of unique people who had been tested, while some states defined their tests as the total number of tests conducted, which included counting repeat testers multiple times (Simon 2021) . Virginia and Georgia initially combined results from diagnostic and antibody tests and then changed their reporting metrics (Kettl 2020; Tahir and Cancryn 2020). In terms of their own metrics, states are now revising their death counts as they have discovered data discrepancies while reviewing death certificates (Simon 2021) .

There have also been numerous discrepancies with regard to reporting race and ethnicity data for COVID-19 testing, cases, hospitalizations, and death counts (National Academy for State Health Policy 2021; Simon 2021) and with regard to the levels of disaggregation in race and ethnicity data (Goldfarb, 2020) . Several states have failed to adequately capture race and ethnicity in their reporting (Hooper, Nápoles, and Pérez-Stable 2020) . According to the National Academy for State Health Policy (2021), as of July 2, 2021, only seven states report race and ethnicity data for testing. States have also varied considerably in terms of the categories used to report race and ethnicity data (Goldfarb 2020 (Simon 2021, unpaginated) . Other states like Louisiana report race data for seven race categories:

American Indian/Alaska Native, Asian, Black, Native Hawaiian/Pacific Islander, Other, Unknown, and White; in addition to two ethnicity categories: Hispanic/Latino or Non-Hispanic/Latino (National Academy of State Health Policy 2021). Moreover, many states have implemented different approaches for tracking race and ethnicity categories for COVID-19 cases while using different categories for other metrics, such as deaths and hospitalizations (Simon 2021) .

The lack of standard reporting protocols and the differences in tracking and reporting metrics have been further complicated by the lack of transparency in reporting the metrics, either intentional or otherwise. For instance, the key scientist managing Florida's dashboard statistics for the virus was fired, leading to allegations that the state government was attempting to downplay the numbers (Sassoon 2020) . In other states, the data were not reported clearly or were simply not reported. For instance, New York published data about deaths in long-term care (LTC) facilities but not cases; also, even though deaths were reported, there were questions raised about the state's reporting methods and death counts (Villeneuve, Condon, and Sedensky 2021). Arizona did not provide any data about COVID-19 at LTC facilities other than the number of facilities affected; in nine states, no data about COVID-19 cases and deaths in LTC facilities have been available on state dashboards (Simon 2021) . The state cooperation needed to ensure the protection of vulnerable populations in society (Blessett 2015) has been absent. Hence, the lack of data standardization protocols and the fragmented nature of COVID-19 policy responses has resulted in differing degrees of information asymmetries across states both in J o u r n a l P r e -p r o o f regard to general and race/ethnicity data transparency, generating a loss of welfare in areas where such asymmetries are amplified.

In regard to transparency, past research has found that centralized organizations tend to release more data and information than their decentralized counterparts (see Matheus and Janssen 2020), but to our knowledge, the impact of decentralization on data transparency in a federal system has yet to be examined. To understand the factors affecting state policy experimentation and variations-and by extension, the determinants of differential state-level information asymmetries-the following section contains a theoretical discussion of institutions and the policy responsiveness of those institutions to citizen preferences, as these factors comprise key variables in our study. In accordance with agency theory, state actors (agents) should remain responsive to the citizenry (the principal) by reducing information asymmetries through greater transparency (Lane 2005).

To tap state variations in disparities in collecting, tracking, and reporting metrics-thus examining the varying degrees of information asymmetries across states-we use the data collected from the COVID Tracking Project-State Reporting Assessments that was spearheaded by the Atlantic magazine, involving many media organizations (The Atlantic 2021). They collected data over the course of a year, ending March 7, 2021. Along with tracking data on COVID-19 cases, the COVID-19 project analyzed how data was defined, how data was made available, and how data was presented at the state level. They found several variations in each of these categories. In defining data for instance, they found that more than a year into the pandemic, definitions of basic data points, such as what even counts as a COVID-19 case, were still not standardized across states, despite the availability of standardized definitions provided by the Council of State and Territorial Epidemiologists (CSTE). Some states were combining J o u r n a l P r e -p r o o f probable and confirmed counts into streamlined total metrics. In terms of residency factors too, variations existed; some states said that their testing, case, and death counts included only people who live in the state, while counts elsewhere included anyone who was tested within the jurisdiction. In terms of data availability, a number of state variations also exist. For instance, Iowa reported percentages, not raw numbers, such as reporting that Black or African American people account for three percent of the state's COVID-19 cases; however, no information was provided on how many people that three percent represents, nor did the state clarify whether it was three percent of all people with positive test results or rather three percent of positive test results where information about race was provided. In addition to these problems, there were other issues in data reporting, disaggregation, and transparency as noted above; states thus varied widely in their collection and reporting of data.

For assessments of the quality of state data reporting and transparency, the COVID-19 tracking project included publicly available data, along with frequent conversations with jurisdictions to understand the data they shared; this in-depth data collection was used to provide an assessment of state reporting efficacy (The Atlantic 2021). We use their data and categorizations of state reporting efficacy to focus on two dependent variables of interest in this paper: (i) State-level Metrics (SLM): this variable measures how effectively states define and report key metrics, such as testing data, cases, hospitalizations, and deaths; and (ii) State Race and Ethnicity Data (SRED): this variable indicates the level of comprehensiveness and efficacy in state collection and reporting of race and ethnicity data (The Atlantic 2021). The data for the two variables reflects the level of detail of each state's reporting and the completeness and clarity of their data descriptions, thus representing our operationalization of transparency aimed at examining differential state-level information asymmetries. The indicators for both variables are operationalized into three ordered categories: 1= few issues exist, 2 = some issues exist, and 3= serious issues exist (The Atlantic 2021). To illustrate what these categories entail, an example of what they indicate for three states is provided below in Table 1 . (Grossman et al. 2020; Neelon et al. 2021) . Public support for governors during the pandemic vacillated depending on whether they framed COVID-19 as a public health or economic issue, and party incongruences between governors and citizens, coupled with political polarization, have prevented a "rally around the flag" effect (Shino & Binder 2020) . Data-driven responses focused on providing granular statistics on the number of cases, the number of deaths, and hospital capacity were used by leaders across jurisdictions; a reliance on data projections allowed U.S. governors to make policy decisions early in the pandemic "without inciting political propaganda" (Turrini et al. 2020, p. 749) . However, these state leaders have not always leaned on the most pertinent statistics, as was made apparent when some governors cited statewide hospital capacity data to justify loosening restrictions without considering the number of cases and deaths or whether individual hospitals were experiencing an overload (Salter & Tanner, 2020) .

With the hypothesis that Republican governors are more likely than their Democratic counterparts to have data issues due to their tendency to enact less stringent COVID-19 regulations (Adolph et al. 2021b; Baccini and Brodeur 2020; Gusmano et al. 2020 ), a dummy variable is included for gubernatorial partisanship 1 .

The legislative branch has also been important in understanding policy adoptions and variations across states (Sapat 2004; Rabe 2004 Rabe , 2007 Wiener and Koontz 2012 however, after September 2020, Republican states fell behind their Democratic counterparts (Neelon et al. 2021 ). In addition to the legislative partisanship, prior research has shown that the professionalism of legislatures is an important predictor and explanation of state policies (Squire 2007 (Squire , 2017 Jansa et al. 2018 ; Weiner and Koontz 2012).

The literature has also examined the influence of the gender and race/ethnicity of state legislators. Some studies have found that, relative to their male counterparts, female legislators have a more liberal agenda, show a greater commitment to feminist activities, and support issues traditionally seen as important to women, such as welfare, education, and health (Swers 2001). Female legislators have also been found to be more responsive to their constituents compared to their male counterparts (Lowande, Ritchie, and Lauterback 2019) and are likely to pay more attention to data standardization, transparency, and accountability. (Porumbescu 2015) . Other studies, however, reveal mixed results (Reingold 2019). In an analysis of "women-friendly" state policies, Cowell-Meyers and Langbein (2009) found that an increase in women's representation in the legislature predicted the adoption of only five of the 34 policies examined (512) Merrit (2020) found that a lack of minority representation in state and local governments has negatively impacted communities of color during the pandemic, but these claims have yet to be tested empirically in the COVID-19 policy literature.

To measure legislative partisanship, we use the party that controlled the state-level legislature as of March 2021 (State Partisan Composition 2021) with the hypothesis that Democratic legislatures will be less likely to have issues with data collection and reporting metrics than Republican legislatures.

Divided government (states with different parties holding the legislative chambers and governorship) typically reduces the chances of enacting conflictual state policies (Bowling and Ferguson 2001) ; since COVID-19 is a contentious and polarizing issue, it was included as a control 2 . Legislative professionalism 2 A variable with an alternate conceptualization of partisanship similarities and differences between legislatures and governors with the following coding was also analyzed: 1 Democratic trifecta (meaning Democrats control the governorship and both chambers of the state legislature), 0 split control (one party controls the governorship while the other party controls at least one legislative chamber), -1 Republican trifecta (Republicans control the (Squire 2007 (Squire , 2017 was included with the expectation that professional legislatures would have fewer data issues. To include the potential effects of gender and race/ethnicity for the reasons discussed above, the percentage of state female legislators and the percentage of non-white state legislators (NCSL, 2021) were included in the analysis, with the hypothesis being that higher percentages of both would lead to fewer data issues. governorship and both legislative chambers). In the case of Nebraska, which is nonpartisan and unicameral, a -1 was assigned for 2020 as it was functionally controlled by the Republicans (Republican governor and the majority of the members of the Unicameral legislature were overtly Republican). However, this variable on partisanship was not statistically significant and the separate partisanship variables for the legislature, governor, and divided government were better predictors for both dependent variables. Hence the combined partisanship variable was not used in the analysis.

In addition to institutional characteristics, democratic principles lead to the expectation that policymakers (agents) will be responsive to their constituencies (the principal). Past theoretical work has discussed the importance of citizen ideology, public opinion signals, problem severity, and needs (Lax and Phillips 2009; Berry et al. 1998; Sapat 2004; Ringquist 1994; Wouters and Walgrave 2017) .

Citizen ideology within states has received much attention in the literature due to the democratic implications associated with policy responsiveness to this variable (Berry et al. 1998 ).

Policymakers may respond to citizen preferences by reflecting ideological stances that can impact re- Under the assumption that state policymakers and public officials should remain responsive to the preferences of the citizenry to fulfill the goals of democracy (Berry et al. 1998; Lax and Phillips 2009; Ringquist 1994) , we use citizen activism and protests as a signal of ideological preferences to assess the effect of these preferences on the efficacy of state-level data collection and data transparency. Past research has found that "Blue Lives Matter" activism and support for law enforcement since the 2016 election has been associated with greater support from conservatives, while BLM protests are usually supported by liberals (Solomon et al., 2021; Drakulich et al., 2020) . These variables thus provide a different type of signal of ideological preferences that reflects the unique circumstances of pandemicrelated data as noted below, as compared to more traditional measures of ideology (Berry et al. 1998; Jones 2021) . 5

Although the collection, disaggregation, and dissemination of public health data is normally a mundane task, data transparency and usage during the pandemic has become highly politicized.

Throughout the pandemic, states have made claims to data-driven responses to COVID-19, but such data-informed decision-making has been heavily influenced by politics ( 

The variables were then aggregated by state to obtain a raw count of protests within the relevant jurisdictions, with 2019 Census data used to convert the variables into protests per 10,000,000 state citizens (U.S. Census 2020). Given that conservatives are more likely than liberals to support the police and "Blue Lives Matter" (Ekins 2016; Solomon et al. 2021; Updegrove et al. 2020) , the pro-police protests variable represents the signaling of conservative preferences, and the BLM protests variable indicates signaling of liberal ideological preferences. The expectation is that states with more BLM protests/riots per 10,000,000 people will have fewer COVID-19 data issues associated with race/ethnicity relative to states with less activism, while the opposite is expected for pro-police protests.

And since conservatives have been more likely than liberals to downplay the severity of the virus (Pennycook et al. 2020) , we predict the same for all state-level metrics as well.

With respect to responsiveness to needs, prior research has shown that states and local governments are likely to be more responsive to problems in their jurisdictions as they understand local problems more clearly (Rabe 1995 (Rabe , 2004 Nice 1994; Sapat 2004 ). In the COVID-19 context, while the U.S. accounts for 4% of the world's population, by July 2021, it had accounted for 21% of the world's deaths (John Hopkins

University and Medicine n.d.). The abdication of leadership by the executive branch at the federal level left states to fend for themselves (Birkland et al. 2021 ; Kettl 2020) with high case rates and subsequent deaths.

To tap the extent of the problem that states faced, we use the total COVID case rates and deaths as of March 7, 2021. We hypothesize that the effects of higher numbers of cases and deaths are likely to have impacted state data collection and reporting efforts by either prompting them to do more of it (so have fewer issues), or conversely, making it more difficult for them to do so (leading to more serious issues in collecting and reporting data).

Citizen needs are also reflected in the concept of social vulnerability. Social vulnerability, a multidimensional construct captured in socio-economic/human dimensions, pre-disaster inequities, and physical dimensions/exposure, impacts the capacity of households and communities to withstand disasters and their impacts, particularly for the most marginalized segments of society (Cutter et al. 2003; Esnard et al. 2011; Fothergill and Peek 2004; Laska and Morrow 2006; Peacock et al. 2012; Sapat and Esnard 2016; Thomas et al. 2013 Epidemiological Factors (includes data on high risk populations such as elderly adults and individuals with underlying conditions including respiratory conditions, heart conditions, obesity, diabetes, and J o u r n a l P r e -p r o o f conditions related to immunodeficiency); and High-Risk Environments (includes indicators for subgroups of populations whose living or work environments puts them at high risk of contracting the virus, such as nursing home and assisted living residents, prison populations, and workers in high-risk industries-e.g., meat and poultry processing, manufacturing, and passenger ground transportation).

The variables for each of these indices were represented by percentiles, which were then aggregated to form the index. Equal weight is given to each variable within each index; the composite measure for each index is measured on a 0-1 scale (0 = least vulnerable, 1 = most vulnerable) (see Surgo Ventures 2021 for more details) 7 . If states are being responsive to the needs of vulnerable populations, then the impact would be negative-states would be better at data collecting and reporting. However, for each of these variables, it could also be the case that states did not respond effectively, in which case there will be more issues with data collection.

Since our dependent variables are measured at the ordinal level with three categories (1 = few issues exist, 2= some issues exist, 3 = serious issues exist), we fitted an ordinal logistic regression or the proportional odds model, which is used with variables with two or more ordered categories (Menard 2009 ). All models were estimated using STATA statistical software version 15 (StataCorp L.P., College Station, TX).

Table 2 presents descriptive statistics for our dependent and independent variables. The model fit is indicated by the chi-square, which is statistically significant, and the pseudo rsquare of 0.474 9 . With respect to institutional factors, the results support the hypothesis that Democratic controlled legislatures have fewer issues in data efficacy and reporting; the legislative partisanship variable is statistically significant. The negative coefficient for this variable indicates that for a Democratic-controlled legislature, the ordered log-odds of having issues with reporting data decrease by 7.028 while the other variables in the model are held constant. The percentage of female legislators also has a significant effect, but this variable increases the extent of data reporting issues. Variables measuring gubernatorial partisanship and divided government were not statistically significant in this model 10 .

Regarding policy responsiveness to citizen ideological signals, while the BLM variable was not statistically significant, the coefficient for pro-police protests was, supporting the hypothesis that states with higher numbers of pro-police protests were likely to have more issues in reporting data with greater information asymmetries. In other words, a conservative ideological signal predicts an increase or more problems in data transparency. Other statistically significant variables are those measuring the needs and vulnerability of citizens. States with more COVID-19 cases had more issues in reporting data metrics. Similarly, the statistically significant and positive coefficient for the epidemiological factors and minority status and language variables indicates that increases in higher-risk medical populations and minority populations are likely to increase problems in reporting and assessing metrics. However, the vulnerability index for socio-economic status was statistically significant and negative, indicating that a one-unit increase in a state's socio-economic vulnerability index decreased the odds of having some or serious issues. This indicates that states responded to the needs of lower-income populations by improving their efficacy in reporting and reducing information asymmetries. Table 4 presents the results of the analysis for the second dependent variable, state collection and reporting on race and ethnicity data. J o u r n a l P r e -p r o o f agency staff per capita, is supported (statistical significance at the .10 level). As expected, higher state administrative capacity reduces the probability of having more issues in collecting and reporting data.

With respect to policy responsiveness to problems and needs of citizens, an increase in the number of COVID-deaths increased the odds of having more serious issues in collecting and reporting race and ethnicity data, while an increase in the COVID-19 case rates lowered the odds of having issues, indicating better data reporting for race and ethnicity data. None of the citizen ideology signal variables were statistically significant, but vulnerability levels in terms of minority status and language and epidemiological factors were statistically significant and indicate that an increase in the number of minorities and medically higher risk populations increase the odds of a state having some or serious issues in data collection and reporting assessments, confirming the arguments that minority populations are more likely to live in areas with less testing (Sgaier 2020) . Similar to the model for state-level metrics, the socio-economic status variable was statistically significant with a negative coefficient, indicating fewer issues in collecting and reporting race and ethnicity data.

Across both models, our hypotheses with regard to partisanship were supported. These findings confirm our theoretical arguments that the efficacy of data collection and reporting processes, particularly during the pandemic, was influenced to a larger extent by political factors such as partisanship, given the rhetoric used by many Republican governors and legislators to downplay the impact of the virus. Moreover, the stakes were high for both Republicans and Democrats in collecting and reporting data, as case counts and deaths were used to make critical economic and political decisions on lock-downs and openings; given that 2020 was an election year, the stakes were even higher for both political parties to show that they were effectively dealing with the pandemic. Unlike other public health data, COVID-19 health data needed to be reported more frequently, were in a J o u r n a l P r e -p r o o f different format compared to other public health data (for instance, daily counts, moving averages, and dashboards with geospatial data were needed), and were scrutinized more closely by citizens to monitor the situation and assess risks.

However, it seems that legislative partisanship rather than gubernatorial partisanship had a greater influence on data transparency and reducing information asymmetries. Given the active and high-profile media role that governors played during the pandemic with the signing of numerous executive orders (Curley and Federman 2020) and press conferences (Mintrom and O'Connor 2020) , it may seem surprising that governors were not as influential as other factors in this study. For state race and ethnicity data reporting, the results indicate that when governors were important, Democratic governors increased the odds of having greater issues in data reporting and disaggregation. This finding is supported by reports that the push to collect race and ethnicity data even before the pandemic was often initiated by state legislators while sometimes being rejected by Democratic Governors; for instance, a bill passed by the New York state legislature in 2019 to collect disaggregated data on Asian-American groups was vetoed by the Democratic Governor (Zylla and Bernard 2020; Yam 2019) . The analysis in this study also indicates statistically significant associations between state institutional characteristics, such as divided government and state agency capacity, and data collection and transparency practices. This finding supports the contention that governors, while often the public face of the crisis, quickly slipped back into relatively familiar, politically proven policy streams and policy decisions that had been produced over the years (Kettl 2020). As noted above, it was the legislatures in some states that pushed for better data collection efforts; for instance, the Maryland Legislative Black

Caucus pushed the governor to collect and release data related to race and ethnicity to understand the In examining policy responsiveness to citizen preferences-i.e., the accountability of the agent to the principal-it appears that states are more responsive to conservative signals with regard to not reporting data effectively, which may be indicative of state attempts to downplay the pandemic by not being more comprehensive and transparent in their data collection efforts. In responding to problem severity and needs, increasing numbers of cases did increase the likelihood of better reporting of race and ethnicity data; however, increases in COVID-19 case rates did not improve overall data collection and reporting, indicating that states may have responded to the higher number of cases by downplaying the numbers presumably to either open sooner or for political expediency to look like they were handling the situation well. Similarly, the number of COVID-19-related deaths was associated with more issues in reporting race and ethnicity data. This finding may reflect reports that death counts are still being revised and corrected in some states, such as Oklahoma, which is revising its dashboard on death counts (Simon 2021) . Increases in minority and medically higher risk populations also increase the likelihood of having more issues with data metrics and disaggregation, indicating that states did not do a good job of collecting and reporting data on these populations; this supports reporting on potential undercounting of medically vulnerable populations in nursing homes (see, e.g., True et al. 2020 ).

However, it seems that increases in populations with high socio-economic vulnerability do spur better overall collection and reporting of state metrics as well as of race and ethnicity data.

During much of the pandemic, the U.S. response to COVID-19 has been characterized by confusion, intergovernmental friction, a lack of federal-level leadership, contradictory messages, and fragmented policy responses (Benton 2020; Birkland et al. 2021; Kettl 2020) . The responsibility for dealing with the J o u r n a l P r e -p r o o f pandemic and its effects were left to state governments with regard to critical decisions on whether and to what extent to lockdown, how to acquire protective equipment, and how to collect and report data.

State policies varied widely in response to federal devolvement. The lack of testing and state variations in data collection and reporting metrics complicated understandings of the seriousness of the problem and the speed of the spread during the pandemic. Through the lens of agency theory, this paper focuses on understanding these state variations in pandemic-related data collection and reporting and data disaggregation by race and ethnicity. As noted by the American Medical Association (AMA), demographic data is a key tool to fight inequities in COVID-19 pandemic planning and to address health disparities (American Medical Association 2020b). Data specificity is needed not only to understand the impact and reach of the pandemic but also to inform appropriate response and planning (American Medical Association 2020c). Without adequate race and ethnicity data associated with COVID-19 testing, hospitalizations, morbidities, and mortalities, physicians and hospitals will not be able to properly care for their patients (American Medical Association 2020a).

To address this issue, we analyzed predictors of variations across U.S. state policies with regard to collecting and reporting metrics and disaggregating data by race and ethnicity using the theoretical understandings and past research on agency theory, transparency, state institutions, and policy responsiveness. Our findings indicate that state-level information asymmetries with regard to data collection and reporting processes were reduced with Democratic control of state legislatures; however, citizen activism reflecting more conservative policy support, higher numbers of medically vulnerable and minority populations, and female state legislators are associated with more issues in data collection and reporting. Improvements in state-level reporting and disaggregation of race and ethnicity data are seen in Democratic-controlled state legislatures and states with higher COVID case rates, while there is less policy responsiveness with data reporting and disaggregation to higher levels of minority and medically at-risk populations.

Although these factors are important, we acknowledge that the approach and data used in this study too have their limitations. Given the paucity of data, our analysis relies in part on data collected by non-governmental and media organizations. Though it could be argued that these measurements lack validity, past research has used portions of the Atlantic's COVID Tracking Project as covariates in statistical models (see, e.g., Famiglietti and Leibovici 2021; Guo et al. 2021; VoPham et al. 2021) ,

demonstrating the project's utility in academic research. The nature of our dependent variables, which calls for a logistic regression model estimation, also has limitations that occur with smaller sample sizes.

We rely on a cross-sectional analysis; studying the issue of equity measures over time could provide different insights. However, panel data on the variables for each of the 50 states were not possible to attain.

Despite these limitations, our research sheds light on an issue that is important in understanding the actions of governments in addressing health disparities and equity through effective data transparency practices. While there has been research on the role of political actors and the protective actions or otherwise that states have taken to address the pandemic (Adolph et al. 2021b; Baccini and Brodeur 2020; Bowman and McKenzie 2020) , little attention has been paid to the issue of data equity in terms of data collection and reporting as a way to reduce information asymmetries and improve the accountability of agents (public officials) to the principal (the citizenry). This study helps to address that area. Moreover, our emphasis on data transparency and equity is critical in examining pandemic responses, particularly since states used the data on cases, test positivity rates, and hospitalizations when needed to justify restrictions, lockdowns, and other mitigation and adaptation policies. While data surveillance and collection can be intrusive, collecting key data, particularly for vulnerable populations, and communicating risk based on that data was successful in other countries, such as South Korea, leading to increased legitimacy for government actions and a 'trust surplus' with increased levels of trust J o u r n a l P r e -p r o o f in government; this led to improved health outcomes with lower case rates, hospitalizations, and death rates (Moon 2020).

The policy implications from our study suggest that there is a growing and critical need for standardized data collection protocols, collection of comprehensive race and ethnicity surveillance data, and data repositories. The salience of such implications extends beyond the United States, as other countries have experienced racial and ethnic COVID-related disparities (Yaya et al. 2020) , and methods for counting cases and deaths have been country-specific, thus lacking cross-national comparability and global uniformity (Karanikolos and McKee 2020). Indeed, an interrogation of official death tolls via an examination of excess deaths across countries displays an undercount of COVID-19 mortality rates (The Economist 2021). Moreover, other federalist systems may have the same information asymmetries and data transparency issues as those in the United States, where the devolution of authority and the absence of a consistent federal message have engendered deleterious health consequences for state populations (Kettl, 2020 (Kettl, , 2021 .

What specific steps can policy-makers and practitioners take to improve these processes? We advance four suggestions here 11 :

First, policy-makers, public health officials, and emergency managers within the U.S. and globally should be aware of the essential role they play in appropriately collecting and disseminating public health data and the importance of mitigating health disparities to increase social equity outcomes for marginalized and vulnerable communities. Second, state and local governments and organizations, including schools and universities, need to be transparent about health data and maintain and disseminate public health data via mechanisms such as public health dashboards (that include clear definitions of the data and other information) and through other avenues to be easily accessible. Data 11 Also noted in Campbell et al. 2021. J o u r n a l P r e -p r o o f definitions need to be accessible from a state's data pages and presented in easily and intuitively accessible locations (The Atlantic, 2021). Third, data dissemination of public health data during pandemics and other disasters needs to be high-quality, trustworthy, reliable, timely, frequent, and accurately reflect changing statistics. Finally, data needs to be disaggregated by various factors such as ethnicity, race, age, geographic area, and other socioeconomic variables to understand the impacts of pandemics and disasters while discerning how to appropriately protect vulnerable populations.

The rise of new variants and continuing waves of increasing infections in different areas of the world make it likely that the COVID-19 pandemic will evolve and likely turn endemic. Lessons learned from this pandemic in terms of data collection, transparency, and dissemination will be critical in reducing disaster risks and improving resilience to other pandemics and disasters in the future.

We also ran analyses using more traditional measures of citizen ideology. Since the most recent Berry et al. measure of citizen ideology is from 2016 in the file on Fording's website (Fording n.d.) , we used instead a measure of citizen conservative and liberal identification taken from a Gallup US poll conducted in 2018 (Jones 2021) . The variables from this data include the percentage of respondents identifying as conservative, the percentage of respondents identifying as liberal, and conservative advantage (measuring the percentage point difference between the conservative and liberal Identification percentages). While the variable 'conservative advantage' was statistically significant in the State Level Metrics model at the 10% level, it lowered the predictive power of the model; specifically, the explanatory power of the model was nearly halved when including the variable conservative advantage in lieu of the signaling of citizen ideology via the indicators measuring levels of political protest (both pro-police and pro-BLM). The results of that model are presented in Table A .1 below. In further analyses, the variables measuring the percentage of respondents identifying as conservative and liberal in each state were not supported in any of the models. Similarly, analyses utilizing Berry et al.'s operationalization of citizen ideology returned statistically insignificant results, high multicollinearity, and a substantially weaker model. 

An Act addressing COVID-19 data collection and disparities in treatment, MA H.4672, 192 nd General Courts of the Commonwealth of

Armed Conflict Location & Event Data Project (ACLED)

Truth or Dare? Detecting Systematic Manipulation of COVID-19 Statistics

The Pandemic Policy U-Turn

Public Health, and Race in Decisions to Ease COVID-19 Social Distancing Policies in the United States

Pandemic Politics: Timing State-Level Social Distancing Responses to COVID-19

Protest and Institutional Change

COVID-19 and Symbolic Action: Global Pandemic as Code, Narrative, and Cultural Performance

COVID-19 FAQs: Health Equity in a Pandemic

The Role of Data Collection in the COVID-19 Pandemic

States Tracking COVID-19 Race and Ethnicity Data

The COVID-19 Pandemic as a Game Changer for Public Administration and Leadership? The Need for Robust Governance Responses to Turbulent Problems

How the Coronavirus Pandemic Helped the Floyd Protests Become the Biggest in US History

Explaining Governors' Response to the COVID-19 Pandemic in the United States

The Politics of Need: Examining Governors' Decisions to Oppose the "Obamacare" Medicaid Expansion

Intergovernmental Relations in the Early Twenty-first Century: Lingering Images of Earlier Phases and Emergence of a New Phase

Challenges to Federalism and Intergovernmental Relations and Takeaways Amid the COVID-19 Experience

Measuring Citizen and Government Ideology in the American States, 1960-93

Freedom of Information and Openness: Fundamental Human Rights?

Governing in a Polarized Era: Federalism and the Response of US State and Federal Governments to the COVID-19 Pandemic

Disenfranchisement: Historical Underpinnings and Contemporary Manifestations

Organising for Effective Emergency Management: Lessons from Research

Leadership in Times of Crisis: A Framework for Assessment

The Politics of Crisis Management: Public Leadership Under Pressure

Divided Government, Interest Representation, and Policy Differences: Competing Explanations of Gridlock in the Fifty States

Fragmented Federalism: The State of American Federalism 2012-13

Managing a Pandemic at a Less than Global Scale: Governors Take the Lead

Do politicians racially discriminate against constituents? A field experiment on state legislators

California Department of Public Health

Temporal Clustering of Disorder Events During the Covid-19 Pandemic

Incorporating Lessons Learned during COVID-19 into Future Planning for Emergencies and Disasters

Policy Preferences and Policy Change: Dy-namic Responsiveness in the American States

Demographic Trends of COVID-19 Cases and Deaths in the U.S. Reported to the CDC

Trends in Number of COVID-19 Cases and Deaths in the U.S. Reported to CDC, by State/Territory

American Federalism in an Era of Partisan Polarization: The Intergovernmental Paradox of Obama's 'new nationalism'

Ethics of Research at the Intersection of COVID-19 and Black Lives Matter: A Call to Action

25 years of Transparency Research: Evidence and Future Directions

State Executive Orders: Nuance in Restrictions, Revealing Suspensions, and Decisions to Enforce

The Effect of the 2020 Racial Justice Protests on Attitudes and Preferences in Rural and Urban America

Social Vulnerability to Environmental Hazards

Federalism in a Time of Coronavirus: The Trump Administration, Intergovernmental Relations, and the Fraying Social Compact

Local Government Transparency Index: Determinants of Municipalities' Rankings

When Does Transparency Generate Legitimacy? Experimenting on a Context-bound Relationship

Introduction: The Legitimising Effect of Social and Environmental Disclosures-a Theoretical Foundation

Race and policing in the 2016 presidential election: Black lives matter, the police, and dog whistle politics

Tracking COVID-19 Excess Deaths Across Countries

An Index of Relative Displacement Risk to Hurricanes

The Impact of Health and Economic Policies on the Spread of COVID-19 and Economic Activity (FRB St. Louis Working Paper 2021-005)

US COVID-19 State Government Public Dashboards: An Expert Review

State Ideology Data

Poverty and Disasters in the United States: A Review of Recent Sociological Findings

Infotopia: Unleashing the Democratic Power of Transparency

Social Vulnerability and Equity: The Disproportionate Impact of COVID-19

Protest and Congressional Behavior: Assessing Racial and Ethnic Minority Protests in the District

The Impact of Protest on Elections in the United States

The State of COVID-19 Race and Ethnicity Data

Political Partisanship Influences Behavioral Responses to Governors' Recommendations for COVID-19 Prevention in the United States

Containment: An Observational Study in the United States

Partisanship in Initial State Responses to the COVID-19 Pandemic

Data, Politics and Public Health: COVID-19 Data-Driven Decision Making in Public Discourse

In Transparency: The Key to Better Governance?

COVID-19 and Racial/Ethnic Disparities

Do Pandemics Lead to Rebellion? Policy Responses to COVID-19, Inequality, and Protests in the USA (WIDER Working Paper No. 2021/57)

Copy and Paste Lawmaking: Legislative Professionalism and Policy Reinvention in the States

Corona Virus Resource Center

Conservatives greatly outnumber Liberals in 19

Leadership of Mayors and Governors During Crises: A Systematic Review on Tasks and Effectiveness

Social Vulnerabilities and Hurricane Katrina: An Unnatural Disaster in New Orleans

Determinants of Voluntary Internet Financial Reporting by Local Government Authorities

Gay Rights in the States: Public Opinion and Policy Responsiveness

Salient Target populations and the Subcategorization of Deviants in the Release of Inmates During the COVID-19 Pandemic

Descriptive and Substantive Representation in Congress: Evidence from 80,000 Congressional Inquiries

What Factors Predict the Passage of State-Level E-Cigarette Regulations?

The CDC and States Are Misreporting COVID-19 Test Data

Politics of Persuasion

It's Just Cuckoo': State's Latest Data Mishap Causes Critics to Cry Foul

Framing a Needed Discourse on Health Disparities and Social Inequities: Drawing Lessons from a Pandemic

A Systematic Literature Study to Unravel Transparency Enabled by Open Government Data: The Window Theory

Opinion | The Pandemic's Missing Data. The New York Times

Tensions in State-Local Intergovernmental Response to Emergencies: The Case of COVID-19

COVID-19 and Health Disparities: The Reality of 'The Great Equalizer'

Logistic Regression: From Introductory to Advanced Concepts and Applications

The Importance of Policy Narrative: Effective Government Responses to Covid-19

Fighting COVID -19 with Agility, Transparency, and Participation: Wicked Policy Problems and New Governance Challenges

Maryland Legislative Black Caucus Pushes for State to Release Racial Breakdown of Coronavirus Impact

How States Collect, Report, and Act on COVID-19 Race and Ethnicity Data

National Public Radio. 2020

Associations Between Governor Political Affiliation and COVID-19 Cases, Deaths, and Testing in the U.S

Policy Innovation in State Government

Social Vulnerability and Hurricane Ike: Using Social Vulnerability Mapping to Enhance Coastal Community

After Ike: Severe Storm Prediction, Impact, and Recovery on the Texas Gulf Coast

Predictors of Attitudes and Misperceptions About COVID-19 in Canada, the U.K., and the USA

Citizen Attitudes Toward Transparency in Local Government

Using transparency to enhance responsiveness and trust in local government: can it work?

Government Ideology and Economic Policy-Making in the United States-a Survey

Social Equity in the Trump Era: What Can Local Public Administrators do to Improve Social Equity for their Residents and Community in the Face of Federal Cuts?

Republicans Respond to Black Lives Matter with Anti-Protest Bills

Integrated Environmental Permitting: Experience and Innovation at the State Level

Statehouse and Greenhouse: The Emerging Politics of American Climate Change Policy

Race to the Top: The Expanding Role of U.S. State Renewable Portfolio Standards

Introducing ACLED-Armed Conflict Location and Event Data

Policy Influence and Policy Responsiveness in State Pollution Control

As Virus Grows, Governors Rely on Misleading Hospital Data

Devolution and Innovation: The Adoption of State Environmental Policy Innovations by Administrative Agencies

Coming home After Disaster: Multiple Dimensions of Housing Recovery

Florida Scientist was Fired for 'Refusing to Manipulate' COVID-19 Data, She Said

Women's Leadership is Associated with Fewer Deaths During the COVID-19 Crisis: Quantitative and Qualitative Analyses of United States Governors

With COVID-19, Rural Testing Deserts Should be our Immediate Priority

Expressions of American White Ethnonationalism in Support for 'Blue Lives Matter'

Equity

The Contributions of the Economics of Information to Twentieth Century Economics

Excess Deaths During the COVID-19 Pandemic: Implications for US Death Investigation Systems

Vulnerable Communities and COVID-19: The Damage Done, and the Way Forward

Defying the Rally During COVID-19 Pandemic: A Regression Discontinuity Approach

Analysis & Updates: Inconsistent Reporting Practices Hampered our Ability to Analyze COVID-19 Data. Here are Three Common Problems we Identified

Measuring State Legislative Professionalism: The Squire Index Revisited

A Squire Index Update

Dynamic Representation

Bad State Data Hides Coronavirus Threat as Trump Pushes Reopening

Determinants of Local Governments' Transparency in Times of Crisis: Evidence from Municipality-level Panel Data

Early CDC Coronavirus Test Came With Inconsistent Instructions And Cost The U.S. Weeks

State Reporting Assessments

Social Vulnerability to Disasters

Black Lives Matter… Still: Moving Beyond Acknowledging the Problem Toward Effective Solutions in Graduate Training and Education

Overlooked and Undercounted: The Growing Impact of COVID-19 on Assisted Living Facilities. KFF

Sense or Sensibility? Different Approaches to Cope With the COVID-19 Pandemic

State Population Totals and Components of Change

2019 ASPEP Datasets & Tables

QuickFacts New Hampshire

Red States and Black Lives: Applying the Racial Threat Hypothesis to the Black Lives Matter Movement

Performing Rituals of Affliction: How a Governor's Press Conferences Provided Mediatized Sanctuary in Ohio

Social Distancing Associations with COVID-19 Infection and Mortality are Modified by Crowding and Socioeconomic Status

Agenda Seeding: How 1960s Black Protests Moved Elites, Public Opinion and Voting

Extent and Types of Small-Scale Wind Policies in the U.S. States: Adoption and Effectiveness

A 'Glorious Poetic Rage'

Demonstrating Power: How Protest Persuades Political Representatives

Social Equity and COVID-19: The Case of African Americans

How the United States Flunked the COVID-19 Test: Some Observations and Several Lessons

Gov. Cuomo vetoes bill to collect demographic data on Asian American groups

Ethnic and Racial Disparities in COVID-19-Related Deaths: Counting the Trees, Hiding the Forest

Advances in States' Reporting of COVID-19 Health Equity Data

Acknowledgements: This work is based on research supported by a U.S. National Science Foundation Grant, (SES-2028958). The findings are not necessarily endorsed by the NSF. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

+/-0.056** 0.025 1.057 0.027 J o u r n a l P r e -p r o o f

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:J o u r n a l P r e -p r o o f