Social Interactions and the Optimal Design of Welfare Systems
Lead Research Organisation:
Institute for Fiscal Studies
Department Name: IFS Research Team
Abstract
Regarding the provision of public services, whether in education, health, poverty alleviation, or others, quality of service is as important as quantity in both developed and developing nations. Those services are provided by teachers, doctors, nurses, or social workers (for example), and discussing quality of services means discussing their recruitment and performance.
Our project studies this within the welfare system. We focus on the importance of social worker (SW) quality for the families they serve, and on how to redesign the allocation of SWs to families to improve the overall effectiveness of the system.
We examine a social programme, Chile Solidario (CS), targeting the poorest 2.5% Chileans, and generalise the lessons learned to other contexts. We also provide new statistical and economic tools to address important and challenging issues in our analysis: how to measure SW quality, how to account for social interactions in the choices and behaviours of poor families, how to optimally allocate SWs to families with or without the presence of social interactions across families.
1. What difference do SWs make to the outcomes of the families they serve? What determines their effectiveness? How important is the selection of SWs for each family?
Most studies evaluating welfare services ignore the provision quality, mainly because quality is difficult to measure. Yet, quality differences between providers may drive large differences in their impact. Because of the difficulty in collecting relevant, observable input-based quality measures, researchers have turned to performance-based measures. The most glaring example is education, where teacher value added (VA) estimates have gained traction as a measure of teacher quality. Teacher VA measures the average learning of children taught by a particular teacher. In this project we will produce the first estimates of SW VA for a Chilean welfare program targeting the poorest members of society (bottom 2.5% of the income distribution). We recognise that SWs operate on several dimensions of families' lives. Therefore, we will compute VA in terms of different outcomes and will assess the extent to which SWs specialise across outcomes.
2. How large are spillovers in social program participation? Through which mechanisms do they operate? How do spillovers affect the design and targeting of social programs?
Typical social programme evaluations focus on individual impacts. There is however increasing concern that social programs have impacts well beyond the individual. Possible spillovers from treated individuals to other members of the community have been increasingly added into program evaluations, and even influenced their designs. It is particularly fitting to examine spillovers in the context of anti-poverty programs. Social interactions may influence both the take up of such programs (say, through the potential of stigma, or through the sharing of information), and their impacts on the lives of the poor. The empirical models studying such spillovers are relatively limited though, so far ignoring feedback effects generated by interactions between members of the same group. Feedback effects are interesting because they can amplify or dampen programme impacts, possibly leading to participation and non-participation traps. We build on recent econometric work on these topics to identify and estimate models with social interactions.
3. How do we quantify SW VA in the presence of social interactions? How do interactions change the optimal allocation of SWs to families?
It is natural to put these questions together and re-examine the calculation of SW VA in the presence of spillovers. In this context, the impact of a SW can go well beyond its impact on the outcomes of the targeted family. Social interactions of this type can affect the optimal match of families and SW. What matters is not only this match, but also the way it spreads through the network.
Our project studies this within the welfare system. We focus on the importance of social worker (SW) quality for the families they serve, and on how to redesign the allocation of SWs to families to improve the overall effectiveness of the system.
We examine a social programme, Chile Solidario (CS), targeting the poorest 2.5% Chileans, and generalise the lessons learned to other contexts. We also provide new statistical and economic tools to address important and challenging issues in our analysis: how to measure SW quality, how to account for social interactions in the choices and behaviours of poor families, how to optimally allocate SWs to families with or without the presence of social interactions across families.
1. What difference do SWs make to the outcomes of the families they serve? What determines their effectiveness? How important is the selection of SWs for each family?
Most studies evaluating welfare services ignore the provision quality, mainly because quality is difficult to measure. Yet, quality differences between providers may drive large differences in their impact. Because of the difficulty in collecting relevant, observable input-based quality measures, researchers have turned to performance-based measures. The most glaring example is education, where teacher value added (VA) estimates have gained traction as a measure of teacher quality. Teacher VA measures the average learning of children taught by a particular teacher. In this project we will produce the first estimates of SW VA for a Chilean welfare program targeting the poorest members of society (bottom 2.5% of the income distribution). We recognise that SWs operate on several dimensions of families' lives. Therefore, we will compute VA in terms of different outcomes and will assess the extent to which SWs specialise across outcomes.
2. How large are spillovers in social program participation? Through which mechanisms do they operate? How do spillovers affect the design and targeting of social programs?
Typical social programme evaluations focus on individual impacts. There is however increasing concern that social programs have impacts well beyond the individual. Possible spillovers from treated individuals to other members of the community have been increasingly added into program evaluations, and even influenced their designs. It is particularly fitting to examine spillovers in the context of anti-poverty programs. Social interactions may influence both the take up of such programs (say, through the potential of stigma, or through the sharing of information), and their impacts on the lives of the poor. The empirical models studying such spillovers are relatively limited though, so far ignoring feedback effects generated by interactions between members of the same group. Feedback effects are interesting because they can amplify or dampen programme impacts, possibly leading to participation and non-participation traps. We build on recent econometric work on these topics to identify and estimate models with social interactions.
3. How do we quantify SW VA in the presence of social interactions? How do interactions change the optimal allocation of SWs to families?
It is natural to put these questions together and re-examine the calculation of SW VA in the presence of spillovers. In this context, the impact of a SW can go well beyond its impact on the outcomes of the targeted family. Social interactions of this type can affect the optimal match of families and SW. What matters is not only this match, but also the way it spreads through the network.
Planned Impact
We intend to contribute to four groups in society: (1) public policy makers; (2) providers of welfare programmes; (3) social workers; and (4) the recipients of welfare services.
1) Public Policy Makers
There is widespread concern with the standards of delivery of social work across locations. This motivated the establishment of the Social Work Task Force by the previous government, which published several recommendations regarding (among other things) training and recruitment of social workers.
However, output measurement is notoriously difficult in the public sector, which impairs the design of human resources policy. Without being able to document the extent to which quality varies across social workers (SW), and what determines this variation, how can one think about the recruitment, retention, compensation, or training of SWs? The research in this proposal is intended to contribute to closing this knowledge gap, by providing the first measures of SW value added for a fairly developed welfare system, and studying the consequences of using different mechanisms for allocating SWs to families.
2) Providers of Welfare Programmes
While national policy makers can offer guidance and rules, many day to day and longer-term human resource decisions are made by local providers. Across the world, these can be in the public or the non-profit sectors. At the very least, local providers need to be able to measure service quality, and should aim to match each SW, with her specific bundle of strengths and weaknesses, to the family who needs that bundle of qualities the most. In contexts of scarce resources, service providers also need to know which families to target.
Our research addresses all these issues. It proposes a procedure for measuring SW quality and presents estimates from such a procedure. It develops a method to estimate impacts of social programs not only on the targeted family, but on its entire social network. This information is important when one thinks about the targeting of services. Finally, we provide and apply algorithms for allocating SWs to families when SWs have qualities especially suited for dealing for specific types of families, and when there are important social interactions between programme recipients and their neighbours.
3) Social Workers and their representatives
Our research focuses on the behaviour of SWs and their impact on the families they service. While SWs have in-depth knowledge about the families they support, they typically lack access to data from the whole welfare system, and are unlikely to be able to extract lessons from the academic literature as readily as academic researchers. We will interact with representative bodies of SWs in the UK and abroad so they both input, and benefit from, our work.
4) Users of welfare services
The end users of the public services delivery we study are poor families. Our contribution to this group will primarily come indirectly by improving the policies of government and local providers of welfare services, and communicating with national bodies representing staff.
1) Public Policy Makers
There is widespread concern with the standards of delivery of social work across locations. This motivated the establishment of the Social Work Task Force by the previous government, which published several recommendations regarding (among other things) training and recruitment of social workers.
However, output measurement is notoriously difficult in the public sector, which impairs the design of human resources policy. Without being able to document the extent to which quality varies across social workers (SW), and what determines this variation, how can one think about the recruitment, retention, compensation, or training of SWs? The research in this proposal is intended to contribute to closing this knowledge gap, by providing the first measures of SW value added for a fairly developed welfare system, and studying the consequences of using different mechanisms for allocating SWs to families.
2) Providers of Welfare Programmes
While national policy makers can offer guidance and rules, many day to day and longer-term human resource decisions are made by local providers. Across the world, these can be in the public or the non-profit sectors. At the very least, local providers need to be able to measure service quality, and should aim to match each SW, with her specific bundle of strengths and weaknesses, to the family who needs that bundle of qualities the most. In contexts of scarce resources, service providers also need to know which families to target.
Our research addresses all these issues. It proposes a procedure for measuring SW quality and presents estimates from such a procedure. It develops a method to estimate impacts of social programs not only on the targeted family, but on its entire social network. This information is important when one thinks about the targeting of services. Finally, we provide and apply algorithms for allocating SWs to families when SWs have qualities especially suited for dealing for specific types of families, and when there are important social interactions between programme recipients and their neighbours.
3) Social Workers and their representatives
Our research focuses on the behaviour of SWs and their impact on the families they service. While SWs have in-depth knowledge about the families they support, they typically lack access to data from the whole welfare system, and are unlikely to be able to extract lessons from the academic literature as readily as academic researchers. We will interact with representative bodies of SWs in the UK and abroad so they both input, and benefit from, our work.
4) Users of welfare services
The end users of the public services delivery we study are poor families. Our contribution to this group will primarily come indirectly by improving the policies of government and local providers of welfare services, and communicating with national bodies representing staff.
| Description | This research is still in progress after delays due to the pandemic and turnover in the research team. During this period we have made progress on two components of the grant: 1) quantify differences in the quality of social workers in Chile; 2) estimate social spillovers in the take-up of social programs. 1) We have estimated simple models for the value added by a social worker using administrative records from 2000 to 2009. We consider two main outcomes: the proportion of minimum conditions satisfied (a list of minimum living conditions that should exist in a household) and the CAS score (a means test score which is used to target social programs) of each household two years after being matched to the social worker. We uncover substantial heterogeneity in the effectiveness of social workers, which is promising for the subsequent stages of the award. 2) We started by replicating previous work (Carneiro et al, 2019) documenting that Chile Solidario, CS, impacted the take-up of a subsidy program, Subsidio Unico Familiar, SUF. We use a regression discontinuity design where being assigned to CS depends on a cutoff of the CAS score referred to above and assignment to the programme is plausibly "random" around this cutoff. Participation in SUF drops significantly around the CS cutoff as expected and documented previously. Then we estimated whether the proportion of one's neighbours taking up SUF also jumps around one's CAS at the CS eligibility cutoff. This is interpreted as saying that one's CS participation affects one's neighbours take-up of SUF. We have then proceeded to estimate a simple more structured model of social interactions in program participation and we characterising differential responses according to household characteristics. Our initial estimates are that these interactions are indeed important. |
| Exploitation Route | The outcomes of this funding will inform policymakers and researchers interested in exploiting the spillovers in social programme participation broadly. |
| Sectors | Communities and Social Services/Policy |