A Reminder to Pay Less for Healthcare: Take-up of Increased Reimbursement in a largescale randomized field experiment

Transcription

1 A Reminder to Pay Less for Healthcare: Take-up of Increased Reimbursement in a largescale randomized field experiment Tim Goedemé, PhD Centrum voor Sociaal Beleid Herman Deleeck Studiedag Nudging, Brussels, 21 November 2017 TAKE Reducing Poverty Through Improving the Take Up of Social Policies q

2 Paper Van Gestel, R., Goedemé, T., Janssens, J., Lefevere, E. and Lemkens, R. (2017), A Reminder to Pay Less for Healthcare: take-up of Increased Reimbursement in a large-scale randomized field experiment, CSB Working Paper 17/12, Antwerpen: Herman Deleeck Centre for Social Policy, University of Antwerp. Centrumvoorsociaalbeleid.be ; takeproject.wordpress.com 1

3 Outline Background of this study Methods and data Results Discussion and next steps 2

4 The TAKE project To what extent can we improve poverty reduction by improving the take-up of social benefits? Size, characteristics, determinants and solutions for non-take-up Institutional data, survey + admin data + RCTs Belspo, Universiteit Antwerpen (coördinator), Université de Liège, Federaal Planbureau Bureau du Plan, FOD Sociale zekerheid SPF Sécurité Sociale 3

5 This study When benefits have to be claimed, non-take-up may undermine effectiveness of social policies Limited information, procrastination, do not fully understand costs and benefits of social programs, Simple behavioural incentives might help Evidence on effect of letters on take up of health-related benefits is rather scarce (esp. in RCT setting) Does mailing (letter + flyer) improve the take up of the increased reimbursement of health care costs in Belgium? 4

6 Increased Reimbursement Increased reimbursement (IR) - Main safety net in health care - Reduced co-payments + third party payer + other financial advantages - Some groups automatic (e.g. social assistance beneficiaries) - Income test for others, taken at local health insurance office - Believed to be quite some non-take-up 5

7 Proactive flux and CM Royal decree of January 15th 2014 Rough check with tax register data ( proactive flux ) => selection of potentially eligible low-income households Christelijke Mutualiteit, biggest insurer Large-scale field experiment in collaboration with CM Randomisation of timing of contact 6

8 Research questions What is the effect of non-intrusive incentives on the take up of the IR? - Does the effect last long in time? - Who benefits most from the mailing? - Are there important regional patterns? 7

9 Treatment 10 (out of 19) regional funds 1/6 households: first Letter + flyer Some regional funds carried out additional phone calls - May reduce information costs - May reduce process costs (immediate appointment) - May reduce stigma Language issues 8

10 Only household heads born before 1985 Allocation to groups Random allocation to six groups, stratified by region, postcode sector and age of household head - weights Unit of selection: address, unless more than 5 household heads on same address 9

11 Quality of the randomisation Check randomisation (90% CIs) Maximum Billing Number of Adults Number of Family One parent HH Man Check randomisation (90% CIs) Days Sickness Days Unemployed Days in Psychiatric Days in General Hospital Year of Birth Check randomisation (90% CIs) Daily Defined Doses (DDD) Expenditures by household Expenditures Health Insurer -0,05 0,00 0,05 W6-W3 W6-W2 W6-W1-4,00-2,000,00 2,00 4,00 W6-W3 W6-W2 W6-W W6-W3 W6-W2 W6-W1 10

12 Total effect % of applications and take up of IR (left) and approval rate (Right) Applied Approved Approval rate

13 Timing of effect Frequency and proportion of households that have applied for/ have been awarded IR across time (Wave 1 vs. Wave 4-6) First Application 0 Frequency Proportion Applied Proportion Applied Days until first application 25nov2015 Date First Applied 24aug2016 Approval 0 Frequency Proportion Approved Proportion Approved Days until approval 25nov2015 Date Approved 24aug

14 Who benefits? Average characteristics of Always takers, treated compliers and never takers (cf. Kowalski, 2016) Aantal dagen ziekte / Invaliditeit in 2015 Leeftijd in 2016 % Man Always takers Treated compliers Never takers 13

15 ZIV uitgaven in 2015 Remgelden 2015 Who benefits? Average characteristics of Always takers, treated compliers and never takers (cf. Kowalski, 2016) Groep 3 Groep 2 Groep 1 Groep 3 Groep 2 Groep Never Takers Treated Compliers Always Takers 14

16 Regional variation 40 Take up of IR by regional sickness fund, wave 1 vs. wave

17 Conclusion Simple reminders can have a substantial effect on the take up of social benefits by low-income groups With lasting effects on improved take up Always takers seem to be the most vulnerable, then come compliers, followed by never takers However, effects differ strongly across the population and regions, and coincides only partially with variations on intervention Up to date information for selection of target group is essential 16

18 Limitations No indicator of effective eligibility, income or migration status Regional variation in treatment and characteristics of population Very specific low-income population 17

19 Recommendations Should definitely be repeated Time between receiving fiscal data and proactive contacts should be minimised Should work on proper up-to-date exclusion criteria, which should be tested further Examine further possibilities and effectiveness of targeted measures for hard-to-reach groups Towards a pro-active welfare state? 18

20 Next steps Expansion to younger household heads Experiment to test exact contribution of phone calls We will assess the effect of IR on health expenditures Aim to measure association between receiving flyers and phone calls as well as uptake of IR with information costs, process costs and stigma (online survey) More detailed characteristics of NTU through more extensive personal interviews (TAKE survey) 19

21 Data Data were collected in September 2016 Administrative records collected by CM - Demographic variables (sept. 2016) - Health care use (2015) - Maximimum billing (2015) - Absence from work (2015) Data and analysis at the family level 20

22 Who benefits? Cannot directly observe the characteristics of those who take up IR as a result of the intervention Characteristics of compliers: weighted average of all takers in the intervention group and always takers in control group Intervention group Never takers (Treated) compliers Always takers Control group Never takers (Untreated) compliers Always takers 21

23 Regional variation Table 8: Gap between sickness funds - LPM's Variable Gap 1 Gap 2 Gap 3 Gap 4 Constant *** * Intervention *** *** *** *** Fund 1 x intervention *** *** *** *** Fund 2 x intervention *** *** *** *** Fund 3 x intervention *** *** *** *** Fund 4 x intervention ** Fund 5 x intervention *** *** *** *** Fund 6 x intervention *** *** *** *** Fund 7 x intervention *** *** *** *** Fund 9 x intervention *** *** *** *** Fund 10 x intervention *** *** *** *** Fund 11 x intervention ** ** ** ** Socio-Demographic* No Yes Yes Yes Healthcare Use* No No Yes Yes Absence from Work* No No No Yes Observations 53,474 53,474 53,474 53,474 R-squared Note: *** p<0.01, ** p<0.05, * p<0.1. Further, the variable categories. * Socio-Demographic, Healthcare Use and Absence from Work are all interacted with the intervention (see in text). 22