Using Administrative Data for Social Science and Policy

A Long-Term Perspective

Recent research has highlighted the utility of administrative data for understanding the long-term outcomes associated with a variety of interventions. Whereas much policy is necessarily driven by research and evaluations examining relatively short-term outcomes, administrative data can provide a longer-term perspective on the effects of past policies and thus a more accurate account of a policy’s costs and benefits. Beyond following individuals who were exposed to interventions over time, administrative data can provide an important tool for understanding multigenerational cycles of advantage and disadvantage, allowing researchers to trace the descendants of individuals from different backgrounds as well as the multigenerational effects of antipoverty policies and interventions. Administrative data also provide opportunities to connect researcher-collected data (for example, observations of student behavior) with future administrative records to help understand the long-term implications of researchers’ observations (such as longer-term cost outcomes related to early school behavior problems).

Administrative data research has pushed the boundaries of what we know about longer-term temporal processes, highlighting the important implications of understanding longer-term policy effects, as well as the intergenerational transmission of advantage and disadvantage more broadly. Standard intervention research typically ends data collection at the end of the study and provides an evaluation at that point. In some large-scale studies, follow-ups track outcomes for multiple years post-intervention (see, for example, Ludwig et al. 2013). Insofar as participants in interventions continue to be captured in administrative records after the end of the study, administrative records provide opportunities for an efficient approach to understanding the benefits of interventions decades later. Given the importance of understanding the long-term impacts of policies aimed at ameliorating poverty (Bailey et al. 2017), and the goal of changing not just individuals’ current circumstances but also their longer-term trajectories, one of the most exciting features of administrative data is their ability to turn any study into a de facto longitudinal study. Recent work in this vein, for example, has highlighted the adult income effects of having an experienced or effective kindergarten teacher and higher-achieving classmates (Chetty et al. 2011; Chetty, Friedman, and Rockoff 2014a, 2014b), the increases in adult civic participation arising from a psychosocial intervention in elementary school (Holbein 2017), and the benefits of cash transfers to poor families on their children’s educational outcomes, mortality, and income (Aizer, Eli et al. 2016).

Administrative data research has also extended a long tradition of survey-based research highlighting the intergenerational transmission of advantage and disadvantage (Blau and Duncan 1967; Ganzeboom, Treiman, and Ultee 1991). Survey-based research has examined parent-child transmission processes (for example, Campbell et al. 2014; Poulton et al. 2002), and in some cases, grandparents (Wightman and Danziger 2014). Administrative data allow researchers to push back on the time horizon of these studies. Using Swedish registry data, for example, Martin Hällsten examines the effects of not only grandparents, but also great-grandparents (2014). Likewise, work by Hyunjoon Park uses Korean historical registry information to examine the outcomes of ethnically Korean slaves, showing substantial effects on their descendants many generations after the abolition of slavery (2014). Multigenerational data are somewhat more rare in the United States, the Utah Population Database being a notable exception (see, for example, Temby and Smith 2014). A more detailed discussion of the opportunities provided by administrative data for understanding multigenerational processes is available from Xi Song and Cameron Campbell (2017). These data provide not only exciting new opportunities to advance the science of the multigenerational transmission of advantage, but also the long-term costs of stigmatized identities, poverty, and disadvantage. As data infrastructure matures, we expect to see more studies examining how social policies aimed at one generation affect their children and grandchildren (for example, Meghir, Palme, and Schnabel 2012).

Key Sites and Populations in the Production of Inequality

Inequality in opportunities and outcomes across groups and persons is one of the most vexing problems facing contemporary society. These inequalities are often produced in spaces that are difficult to examine using surveys or experiments. In the hiring process, for example, correspondence studies can examine who receives responses from potential employers but cannot help us understand how applicant pools are created, which interviewees receive offers, or whether pay differences exist among those who receive offers. Research on inequality in other institutional spheres such as health care, housing, and higher education faces similar challenges. Inequality in institutions is often shaped by processes that determine who is included and excluded, and how individuals are ranked within an organization. Administrative data on hiring pipelines, performance ratings, promotion decisions, and decisions about termination can provide valuable insights into the decisions made by gatekeepers at key sites regarding entry and exit from organizations, as well as important processes governing intra-organizational inequality regimes. One useful feature of workplace administrative records is that they allow researchers to compare individuals with others who have the same occupation working in the same establishment. Research making such comparisons suggests that much of the wage inequality observed across gender and race is created by sorting processes, as individuals doing the same work for the same employer receive largely similar pay (Petersen and Morgan 1995; Tomaskovic-Devey 1993).

Building on this insight, a long tradition of case studies uses administrative records from company human resource departments to understand inequality in these sorting processes. Trond Petersen and Ishak Saporta argue that in the current institutional context the opportunity for firms to discriminate is greatest at the point of hire (as opposed to discrimination in promotion or termination practices) and show that hiring is where the largest differences between men and women are observed (2004). Roberto Fernandez and his collaborators use human resource data in a series of papers that provide important insights into race and gender inequality in the hiring process; they show, for example, the importance of referrals (Fernandez and Greenberg 2013) and the importance of supply-side adjustments to perceived demand-side constraints (Fernandez and Friedrich 2011; Fernandez and Campero 2017). Recent work in this vein highlights the perhaps surprising egalitarian influence of an executive search firm (Fernandez-Mateo and Fernandez 2016), and in the current double issue Fernandez and Brian Rubineau use their extraordinary hiring data to provide novel analysis of network recruitment efforts and their impact on the gender-based glass ceiling in the biopharma industry (2019).

Beyond the labor market, administrative data from other sources also provide important insights into key sites for generating inequality. Research on NIH funding decisions, for example, uses detailed records to document the existence of gender inequality in the NIH review process (Li 2012). Research examining the criminal justice system uses administrative records from juvenile courts to estimate the effects of juvenile incarceration on later criminal justice and school outcomes (Aizer and Doyle 2015). One of the supposed keys to combating inequality and reaching life success is, of course, education. Although we know that dropping out of high school breeds failure and college graduation brings success, much less clear is the value of post–high school associate degrees, vocational diplomas, certificates, and partial college. In this double issue, ChangHwan Kim and Christopher Tamborini merge school administrative data files with earnings files to examine the long-term earnings that accrue from these post–high school accomplishments (2019).

Janelle Downing and Tim Bruckner use housing foreclosure administrative records and birth records to highlight yet another source of inequality (2019). They show that housing foreclosures (and presumably the stress they cause) contribute to premature births and increase inequality in birth outcomes across race and ethnic groups.

Administrative data also allow us to understand small, often difficult to access, populations that are theoretically important. For example, research using large administrative datasets has shown that millionaire tax flight does not occur at levels that are socially meaningful (Young et al. 2016), and that top earners are increasing isolated from the rest of the population (Godechot 2013). Insofar as many large administrative datasets include information on the whole population, these data allow researchers to examine relatively small and theoretically important groups (for example, those that are hard to capture in a probability sample without an explicit oversample) without compromising representativeness (Liebler, Bhaskar, and Porter 2016). From a local policy perspective, the ability to identify small groups of people is helpful because it allows policymakers to ensure that the policies they implement are having their intended effects for all stakeholders, and to determine where adjustments to existing policies are needed (Howard et al. 2019).

Understanding Individuals in Their Social Contexts

The importance of context is a truism in social science research. From network influences to cultural factors to questions about positional goods, contextual considerations play a profound role in shaping an individual’s outcomes. Despite this, interventions and policies have historically operated from a baseline that presupposes constant universal effects that operate at scale if implemented with fidelity (Dodge 2011). The density of information in administrative data is useful in providing opportunities to examine important sources of heterogeneity (particularly contextual sources, but also individual-level factors), as well as providing opportunities to investigate policies and interventions at scale. Durham Connects, for example, takes advantage of administrative data by assigning all newborns in Durham born on even days to receive a nurse home visit (Dodge et al. 2014). This design allows for children to be followed in administrative records throughout their lives in an ethical manner without the requirement of individual consent (because data can be de-identified before being analyzed but retain the essential characteristic of assignment to intervention) without necessitating additional data collection.

Further, thinking about changes to the social system more broadly (for example, moving bus stops to help mothers travel to local clinics) shifts research and policy discussions away from a methodological individualism that focuses on the effects of treatments on individuals, and toward considering how programs and policies affect social systems more broadly (see, for example, Denice and Gross 2016). These systems-level approaches are important for both science and policy, as they address an important shortcoming of much social science research. Research often asks what would happen if everything was held constant and only one consideration was changed. This can be instructive, but it fails to take into account the myriad of ways that people and their social worlds are interconnected. Analyses of how community-level policies and interventions shape not only individuals’ outcomes, but also society more generally, allow researchers to capture the complicated feedback loops and spillover effects that occur when interventions and policies operate at scale, providing insight into how policies might change society more broadly (Dodge 2009; Penner et al. 2015). Although in theory such analyses are possible without using administrative data, in practice the existence of administrative records greatly facilitates them.

Research using administrative data can also provide a more complete account of certain aspects of context, including the government services context. For example, Robert Goerge and Emily Wiegand examine the overlap in families’ access of government services across multiple agencies to show how some families are accessing many agencies whereas others seem to be underaccessing resources that might benefit them (2019). Goerge and Wiegand show that these differences vary across geographical locations within Illinois, suggesting that local practice might contribute to, and mitigate, any biases. Lanikque Howard and colleagues examine the relation between parents’ payment of child support and children’s involvement in the child welfare system (2019). Agustina Laurito and colleagues combine multiple administrative datasets to show how school climate and neighborhood crime levels affect student achievement (2019).

Administrative data can also afford studies that are simply not plausible through original data collection. For example, it is difficult to imagine survey data tracking all of the classmates that a student had or all of the co-workers over an employee’s career, but educational administrative data and linked employer-employee datasets often include this information (Abowd, Haltiwanger, and Lane 2004). Administrative records can thus provide information not only about an individual research subject but also about the environment surrounding them. Christopher Candelaria exploits such data in education, where he disentangles the long-term effects of a third-grade teacher, the medium-term effects of middle-school teachers, and the short-term effects of an eighth-grade teacher (2015).

Linking administrative data files across levels also provides innovative opportunities to understand individual behavior in broader context. Elizabeth Ananat and her colleagues, for example, link county-year-level administrative data about community-level job loss with individual student educational administrative data files in order to discover the impact of local economic downturns on student academic progress (2017). Further, as noted previously, administrative data allow for individuals to be placed in a multigenerational familial context. In providing dense coverage of populations, these data allow researchers to examine whether policies had spillover effects (either positive or negative) on those around the targeted populations, and to examine questions around how context moderates the effectiveness of treatment. Heterogeneity in treatment effectiveness is important not only for contributing to scientific understanding regarding the mechanisms through which interventions work, but also because it has important implications for generalizability and scalability (Domina et al. 2016).

An Iterative Policy Design, Implementation, and Evaluation Cycle

A common goal of scientific research is to improve societal outcomes. Social scientists often seek to do this by evaluating and informing existing policies, and it is not uncommon to hear researchers bemoan the lack of policy responsiveness to research. Although obtaining access to administrative data can be time consuming in some contexts, a potential advantage of administrative data analysis in many contexts is that the institution generating the data is also making and implementing the policies being evaluated, so that there is an audience that is positioned to make decisions about practice and policy based on researchers’ findings (Howard et al. 2019). This is particularly true in researcher-practitioner partnerships, where researchers partner with organizations to help them use their administrative data in better ways to answer questions of interest to decision makers and stakeholders.

The research-policy link in social science is often conceptualized as one in which research informs or evaluates policies, but in research-practitioner partnerships, policy implementation and research can have a bidirectional synergistic relationship. These partnerships provide data researchers could not otherwise access. This unusual opportunity holds not only for companies’ human resource data, but also for educational data, where laws protecting the privacy of student data allow data to be shared with organizations conducting research on behalf of educational agencies to help improve instruction. Further, researchers not only have the opportunity to study important policies in real-world settings but can often inform the implementation process. This relationship not only allows for policies that draw on researchers’ expertise, but can also lead to opportunities for better research because researchers can help implement policies in ways that facilitate high-quality evaluations (such as introducing lottery-based assignments for oversubscribed programs or thresholds in assignment scores that enable regression discontinuity-based designs). In contexts that facilitate the timely incorporation of feedback, data collected can be used to inform Bayesian adaptive designs to help improve interventions in real time (Finucane, Martinez, and Cody 2017).

By bringing researchers into the policymaking process, rigorous research becomes part of the iterative process of policy implementation and adjustment and the policy adjustments made in implementation are better captured by research (Howard et al. 2019). This approach can provide both better policy and better research and serves as a model for how to accumulate and incorporate knowledge beyond simply conducting a series of single-policy evaluations. Although shortening the feedback loop among implementation, research, and redesign is likely to be positive overall, one challenge of this linkage is that long-term outcomes by definition cannot be observed quickly, and many policies—particularly those aimed at implementing organizational changes—have effects that take time to emerge or vary across different stages of the implementation cycle (Mills and Wolf 2017; Sun, Penner, and Loeb 2017).

Combining Datasets Across Sources

Although current administrative data research focuses on contemporary data sources, there is a long tradition of using administrative records in archival and historical research (see, for example, Kessler-Harris 1982; Wilde 2004). Research using administrative data has much in common with history and archeology, insofar as it observes the tracks that individuals leave as they move through society and draws lessons from these glimpses into their lives. A key difference is that when records outlast people, opportunities for supplementing and triangulation through interviews, surveys, or ethnography decline, leaving scientists to reconstruct meaning from the traces people have left behind. Although administrative data researchers using contemporary data draw conclusions from the traces left behind in current records in a similar manner, research using contemporary records has the potential to incorporate information directly from individuals through surveys and observations to supplement the data in administrative records.

Given their origin in a particular institutional context, administrative records are typically fragmented, and these data are often not linked to other data that would be useful for research and policy. Hospitals, for example, collect detailed information about patients’ health, schools regularly collect information about student development, and employers often keep records not only about the performance of employees, but also about applicants who were ultimately not offered positions. Although various combinations of these data can provide important insights, they are typically compartmentalized. Likewise, given their origin, administrative records often lack certain kinds of information that are less likely to be collected in these records. For example, information about attitudes, affinities, and motives are not often collected in administrative records. Combining administrative data with records from other sources—either by linking administrative records across sources or by making administrative records available to be linked to data collected via other means—is thus central to building administrative data infrastructure.

Qualitative Research with Administrative Data

Although much of the research using administrative data uses quantitative information, administrative records also contain vast amounts of qualitative information. Archival research using administrative records provides a strong indication of the considerable value of qualitative work using administrative records. Although qualitative social science research using contemporary administrative records is also just beginning to realize its potential, several examples evince the promise of such approaches. Recent qualitative research in medicine, for example, highlights gender differences in the feedback that medical school residents receive (Mueller et al. 2017), and research on online dating profiles underscores how racial boundaries are reinforced not just by racial homogamy, but also by those looking to date across racial lines (Rafalow, Feliciano, and Robnett 2017).

In many administrative contexts, given the scale of textual data, advances in machine coding offer a promising approach to turning rich qualitative data into quantitative data. In this double issue, Emily Penner and her colleagues provide one example of this approach, showcasing how essays submitted as part of teacher applications are correlated with a variety of policy-relevant considerations (2019). The promise of such approaches in researcher-practitioner partnerships is difficult to overstate, because when these organizations begin to leverage their data in the ways that large tech firms do, there would appear to be substantial benefits for both policy and science. With text mining becoming increasingly sophisticated and common, and the growth of software to aid in the transcription, storage, coding and sorting processes in qualitative research, the distinction between quantitative and qualitative research is one that could quickly fade in administrative data research.

Technical Challenges

Important technical challenges remain to constructing administrative data infrastructure. For example, address-based matches are difficult to implement in contexts that lack a well-defined address system (see, for example, Wynn, Reyes, and Caldwell 2011). Likewise, for computationally intensive analyses (for example, some social network analyses) it is currently not practical to conduct analyses that make use of the density of information available at the population level. These and other questions notwithstanding, in our estimation the largest challenges to administrative data are not technical per se but instead technical constraints imposed in response to legal or perceptual considerations.

For example, it is not clear that there is a strong rationale for why researchers need to be in Texas to analyze data from the state of Texas (except that it may be easier to arrest a misuser within state), or that data from Georgia should be allowed to be used on projects only with a collaborator from a university located in Georgia. Nevertheless, such arrangements remain relatively common. Although they are not insurmountable, they do create nontrivial barriers to access and hinder the democratization that researchers generally support in science. That said, given the level of trust required for companies to allow researchers to analyze key intellectual property (Fernandez-Mateo and Fernandez 2016) or for countries to allow outside researchers access to tax data (King et al. 2017), some restrictions to access beyond those governing survey data are warranted. These barriers highlight the point that the most important challenge to successful administrative data scholarship is not the technical nature of data storage or security, but rather, the human and institutional relationships that must be developed and maintained. The relational nature of data access in many cases—such as in long-term researcher-practitioner partnerships—does result in important constraints that are in tension with norms around data-sharing and open science.

One important challenge surrounding administrative data is the lack of consistency regarding which data are collected and how they are collected. Although national surveys typically use standardized measures and best practices for assessing various constructs, information contained in administrative data can be highly variable in terms of coverage and quality. One advantage of working in close partnership with the organizations generating administrative data is that they typically have a deep understanding of how the data are generated and areas where information may be inaccurate or have limited coverage, and can often adjust practices to generate data that are of mutual interest. Working closely with partners on the ground can also help avoid misattributing causal relations. As with most survey data, administrative data sources require that findings be disseminated only as aggregate statistics in order to protect privacy. As far as we are aware, very few cases of researchers infringing on the privacy of individuals using administrative records have been documented. At this point, then, the technical challenges involved in building data infrastructure are largely surmountable, and the larger remaining question is whether political will is strong enough to move forward.

Legal Challenges

Currently, legal constraints affecting administrative data infrastructure focus on balancing the privacy of individuals whose data are contained in the administrative records with the ability of institutions to find answers to their pressing policy questions, which in many cases will enable them to serve better those who are represented in their data. Allowing access to outside researchers working on behalf of the organization can greatly enhance the research capacity of institutions that generate administrative data and provide expertise in areas that might be otherwise difficult to obtain. In this context, analyses of administrative data should address questions of the data owner, presumably in service of either those represented in the data or the broader public. By contrast, scientists argue that science benefits from widespread, democratic access, and that this access can yield new insights that might be broadly beneficial to society, the institutions generating the data, and their stakeholders, even if these benefits might not have been anticipated. Although making administrative data more widely available is likely generally beneficial, it is currently difficult to know how to assess and weigh the benefits from broad access.

Many forms of administrative data are legally protected in ways that limit access. Under the Family Educational Rights and Privacy Act, identifiable educational data in the United States can only be shared with researchers in a limited number of contexts, including cases where the studies will help the schools improve instruction. Similar challenges apply to health information and Health Insurance Portability and Accountability Act regulations. The lack of a well-established administrative data infrastructure means that lawmakers often do not consider the impact of legislation on administrative records. For example, out of concerns regarding administrative records being used for enforcement purposes, California lawmakers sought to enact laws prohibiting data-sharing and initially did not recognize the limitations this would create for researchers and administrative data infrastructure. Presumably a more robust and salient administrative data infrastructure will help in avoiding such issues in the future.

In many ways, legal constraints are a question of political will. On this point, the bipartisan support for administrative data represented by the Murray-Ryan commission and the Foundations for Evidence-Based Policymaking Act is encouraging. One might imagine, for example, that evidence-based policies around education and workforce training programs might benefit from administrative records from schools, even if the resulting study might not help each school improve instruction. Although individual lawmakers may differ on policy priorities, it is encouraging that they agree on the need for better data and analysis to inform them.

Ethical Challenges

Beyond strictly legal questions, there are ethical questions as well. Typically, potential research participants have the choice to opt out of a study. But this is not possible in most research that uses administrative records. Although consideration of informed consent is routine when it comes to whether a participant’s data are used in traditional research designs, administrative records research is often considered to be nonhuman subjects research. To be clear, questions around individuals’ rights vis-à-vis their data are a feature of administrative data more generally and not particular to research. This is apparent when one considers medical records. In approximately half of the states in the United States, physicians or hospitals own patients’ records, and only in New Hampshire do these data belong to patients (in the remaining states data ownership is not clearly defined). It seems unlikely that patients would take issue with research analyzing these records for patterns that might help save their life. Likewise, it seems probable that most people would not object to their records being analyzed for research that might help save the lives of others. Nevertheless, because this research is often not considered to involve human subjects, and these data (outside New Hampshire) do not belong to the individual, it is unclear what rights patients should have to restrict the use of their data in administrative records research.

Historically, the argument has been that the primary potential harm in this research is that of disclosure, or harm to the individual due to a breach of privacy. Some legal scholars suggest that this individualistic perspective may be problematic. In a high-profile example, the Havasupai sued Arizona State University for using existing blood samples in ways not covered by agreements. In discussing this case, Katherine Drabiak-Syed notes that our current legal system is ill equipped to consider issues beyond an individualistic framework, so that harm to a collective group may not be recognized (2010). These questions are perhaps especially salient in the context of Native Americans, where issues over the right to opt out are laden with colonial legacies of ignoring indigenous perspectives and also raise questions of tribal sovereignty. These concerns are likely heightened where blood (or other physical samples) are involved, where research focuses on historically marginalized populations, or when researchers are partnering with data-collecting organizations. The concerns are perhaps somewhat attenuated when looking at historical data (for example, the Dutch Hunger Winter), but the larger point remains relevant for administrative records research.

More broadly, the issue could be conceptualized as whether individuals should have the right to ensure that their data are not used in systems against their wishes. One might imagine, for example, critics of structural racism not wanting their data to be used by companies that might perpetuate racial differences in homeownership through credit scores. But it is difficult not to be complicit when almost everyone is part of the administrative data ecosystem that creates and reproduces these inequalities. This is a feature of our societal data infrastructure and is not specific to research using this infrastructure. Nonetheless, administrative data researchers should be cognizant of these issues, particularly in contexts like researcher-practitioner partnerships where they might influence the kinds of data collected, and where the research being conducted might be used to justify or rationalize practices that may otherwise be seen as problematic.

Perceptual Challenges

Perceptual challenges relevant to administrative data research can be divided into those within the academy and those in the public domain. Within the academy, in many social science disciplines there is a bias against work that is viewed as overly applied. The term evaluation research for example, is sometimes used pejoratively in contrast with pure science, implying that scientific work is somehow contaminated by being useful to society. We argue that whatever the origins of this bias, it is a distinction that has outlived its usefulness, and that supporting human flourishing—both through better understanding the social world in the broadest and most abstract sense, as well as through understanding the implications of the concrete choices that we as a society make—ought to be one of the aims of science. The degree to which these biases are held in any given scientific field varies, suggesting that social science disciplines can learn much from those more engaged in policy. These disciplinary biases are perhaps a space that academics are well positioned to change. Although these norms may be deeply entrenched, they are nonetheless created and maintained by academics, suggesting that we as a community can change them by changing our hiring criteria, tenure and promotion letters, award nominations, and graduate training. We suspect that these perceptual challenges within the academy are decreasing, in part because administrative data allow researchers to address questions that are not only important for real-world applications, but also make fundamental contributions to discipline-specific and transdisciplinary research goals. We believe the proliferation of administrative data research suggests that over the long term, perceptual issues within the academy are likely to become less pronounced.

Beyond the academy, public skepticism about the limits of confidentiality and data protection threaten public support for the use of administrative data. Recent hacking events and misuse of large private data files at Facebook and Cambridge Analytica have shaken public faith in keepers of supposedly private data. The threat goes beyond misuse to include possible political obstruction by groups such as ALEC (American Legislative Exchange Council), which has taken the position that all governmental action should be minimal. The possibilities of misuse by insiders, hacking by outsiders, and opposition by politics will always be present, but we believe the marginal extra risk imposed by bringing researchers into this circle is very low. Researchers are required to be trained and credentialed in the use of sensitive data files, and universities tend to implement cutting-edge technologies in data security. Because of these threats and the public’s vigilance, however, researchers would be wise to understand the treasure that they behold and to be extremely careful in their use of administrative data files.

At the same time that the public is skeptical, bipartisan support is also strong for administrative data science to improve our capacity to maximize the potential of our human resources. Data-sharing can be difficult in contexts marked by suspicion and mistrust, and larger conversations around privacy remain important. Legal protections governing administrative data use thus play central perceptual and scientific roles, as well as being important for ethical reasons (Anderson and Seltzer 2007). We believe that it is incumbent on scientists to help make the case for administrative data research by ensuring that the public benefits from the use of their data. Although in some cases this might mean working closely with policymakers and practitioners generating and using the data, press coverage of novel findings using engaging data visualizations that reach the public more broadly also play an important role in highlighting the utility of these data to the broader public. Wide dissemination of research findings not only helps inform public discourse around important social questions, but also plays an educational role by engaging people’s curiosity and helping them understand how the social world works.