Adverse Selection and an Individual Mandate: When Theory Meets Practice

Transcription

1 Adverse Selection and an Individual Mandate: When Theory Meets Practice Martin Hackmann, Economics Department, Yale University Jonathan Kolstad, Wharton School, University of Pennsylvania and NBER Amanda Kowalski, Economics Department, Yale University and NBER July 26, 2013

2 Adverse Selection and an Individual Mandate in Theory Theory has long held that there is a welfare loss from adverse (or advantageous) selection Akerlof (1970), Rothschild and Stiglitz (1976) Important in health insurance markets Consumers hold private information Regulations restrict price dierentiation by health type One theoretical way to reduce the welfare loss from selection, at the extensive margin, is a mandate that requires both the healthy and the sick to purchase coverage

3 Adverse Selection and an Individual Mandate in Practice Individual mandate is a centerpiece of the Massachusetts health reform of 2006 and the ACA of 2010 (and at the heart of Supreme Court challenges) In this paper, we test for selection in Massachusetts individual insurance market quantify the welfare impact of the individual mandate from the reduction of adverse selection

4 Literature Early literature tested for presence of asymmetric information in insurance markets Chiappori and Salanie (2000) Finkelstein and Poterba (2006) Small but growing literature on the welfare impact of adverse selection in health insurance markets Einav, Finkelstein, Levin (2011) review Bundorf, Levin, and Mahoney (2012)

5 Literature Limitations persist Previous studies focus on welfare impact of adverse selection on the intensive margin (which plan to choose among employer plans), nd small welfare impact Welfare impact plausibly larger on extensive margin (whether to have insurance at all) and in individual market Analysis of the extensive margin speaks more directly to the theoretical literature No empirical variation from mandates

6 This Paper 1 Develop a simple structural model for the individual health insurance market in MA 2 In this framework, we model the individual mandate (tax penalty) as an exogenous shifter in the price for health insurance 3 We use the model to develop a sucient statistics formula, which allows us to calculate the welfare eects of the mandate in terms of a small number of empirical moments 4 We combine data on insurance coverage, premiums, and average costs with a dierence-in-dierences estimation to strategy to estimate these sucient statistics and the welfare implications of the individual mandate

7 Outline 1 Introduction 2 Institutional Background 3 Theoretical Approach 4 Empirical Approach 5 Results 6 Conclusion

8 Institutional Background: Pre-Reform Community Rating Baseline Model: Assume common price for consumers within a plan Guaranteed Issue Regulations Expect selection to be adverse

9 Institutional Background: Reform Individual Mandate: Pay-or-Play: Consumers that don't have health insurance that meets minimum coverage criteria must pay a tax penalty of $1,260 Overview Mandate is particularly important for individuals who do not have access to employer sponsored health insurance Focus analysis on individual market Introduction of health insurance exchange markets Expect lower post-reform markups Commonwealth Care Program Free or subsidized health insurance for lower income residents (up to 300% of FPL)

10 Outline 1 Introduction 2 Institutional Background 3 Theoretical Approach 4 Empirical Approach 5 Results 6 Conclusion

11 Demand Premium D I,0 Insurance

12 Demand Premium D I,0 P,pre A I,pre Insurance

13 Average and Marginal Costs Premium D I,0 P,pre A AC I MC I I,pre Insurance

14 Average and Marginal Costs Premium D I,0 P,pre AC,pre A H AC I MC I I,pre Insurance

15 Ecient Coverage Premium D I,0 P,pre AC,pre A H AC I MC I I,pre I,efficient Insurance

16 Welfare Loss of Adverse Selection Premium D I,0 P,pre AC,pre A H AC I MC I I,pre I,efficient Insurance

17 Individual Mandate: Tax Penalty Premium P,pre AC,pre D I,0 A H D I,Π AC I Π MC I I,pre I,efficient Insurance

18 Post Reform Equilibrium Premium D I,0 D I,Π P,pre A AC,pre P,post AC,post H A' H' AC I Π MC I I,pre I,post Insurance

19 Post Reform Equilibrium Premium D I,0 D I,Π P,pre A AC,pre P,post AC,post H A' H' AC I Π MC I I,pre I,post Insurance

20 Change in Markup Premium D I,0 D I,Π P,pre A AC,pre P,post AC,post H A' H' AC I Π MC I I,pre I,post Insurance

21 Change in Markup Premium D I,0 D I,Π P,pre AC,pre P,post AC,post A H A'' H'' A' H' AC I Π MC I I,pre I,markup I,post Insurance

22 Optimal Tax Penalty Premium D I,0 D I,Π P,pre A AC,pre P,post AC,post H A' H' AC I Π MC I I,pre I,post Insurance

23 Optimal Tax Penalty Premium P,pre D I,0 A D I,Π D I,Π AC,pre P,post AC,post H A' H' AC I Π Π MC I I,pre I,post Insurance

24 Optimal Tax Penalty Premium P,pre D I,0 A D I,Π D I,Π AC,pre P,post AC,post H A' H' AC I Π Π MC I I,pre I,post I,efficient Insurance

25 Sucient Statistics for Welfare Analysis Estimating welfare changes and the optimal penalty requires only estimates on pre-reform levels and changes in Insurance Coverage Premiums Average Costs Example: W full = (P, pre AC, pre ) (I, post I, pre ) (AC, post AC, pre ) (I, pre + (I, post I, pre )) ((P, post π) P, pre ) (I, post I, pre )

26 Sucient Statistics for Welfare Analysis Pre-reform levels and changes in coverage, premiums, and average costs allow us to pin down the structural objects of our empirical analysis Demand curve Average cost curve Why do we need 6 moments to pin down 4 structural parameters (intercept and slope of the demand and AC curve)? Two additional moments identify the pre-reform and the post-reform markup

27 Outline 1 Introduction 2 Institutional Background 3 Theoretical Approach 4 Empirical Approach 5 Results 6 Conclusion

28 Empirical Approach Read the pre-reform coverage, premium, and average cost levels o the data: Average levels in 2004 and 2005 Use a dierence-in-dierences estimation strategy to quantify reform-related changes in coverage, premium, and average costs between the pre- and and the post-reform years Compare changes in Massachusetts with changes in other states between the pre-reform and the post-reform years Controls for common national demand and average cost shocks that may shift the demand/average cost curve

29 Regression Model Y k st = γ k (MA After) st + ρ k 1 (MA During) st + ρ k 2 MA s + ρ k 3 After t + ρ k 4 During t + ρ k 5 (MA After Year) st + ρ k 6 (MA During Year) st + ρ k 7 (MA Year) st + ρ k 8 (After Year) t + ρ k 9 (During Year) t + ρ k 10 Year t + ρ k 11 + ɛ k st where Y st is measured at the state-year level and refers to Coverage Annual premium Annual average costs Approach is conceptually equivalent to estimating the demand and the AC curve via TSLS

30 Data SNL Financial Database: Compiled from National Association of Insurance Commissioner's reports Detailed data at the rm-market-year level on premiums, expenditures, enrollment in member-months Universe of insurers in the individual market Drop insurers that oer Commonwealth Care Plans

31 Data National Health Interview Survey (NHIS) Allows us to express insurance coverage in percentages To match the SNL population we restrict the sample to individuals 18-64, earning more than 300% of FPL to avoid Medicaid expansions Commonwealth Care plan expansion Ensure homogeneous tax penalty

32 Outline 1 Introduction 2 Institutional Background 3 Theoretical Approach 4 Empirical Approach 5 Results 6 Conclusion

33 Impact on Coverage: SNL Health Insurance Coverage in Population Year Non MA MA

34 Impact on Coverage: SNL

35 Impact on Premiums Average Annual Premium Year Non MA MA

36 Impact on Coverage: SNL

37 Impact on Costs Average Annual Claim Expenditures per Person Year Non MA MA

38 Impact on Coverage: SNL

39 Empirical Graph Corresponding to Theory Premium 7 P,pre 5.9 AC,pre P,post AC,post D I,0 A H D I,Π A'' H'' A' H' AC I Π $1,250 2 MC I I,pre 0.7 I,markup 0.84 I,post Insurance

40 Translating Empirical Estimates into Welfare Use estimates to compute full welfare eect: ( ) W full = $5, $5, % ( ) ( $621.4) 70.3% % 95% CI [$142,$646] + 1 ( $1, 143 $1, 250) 19.4% = $ Putting estimates together gives us the welfare change per individual in the individual health insurance market 212,000 people in the non-medicaid individual market Total welfare gain of $93,704,000 per year Total welfare gain equal to 8.4% of insurer medical expenditures

41 Changes in Markup vs. Adverse Selection Changes in health insurance coverage 13.7% because of removal of adverse selection 5.7% because of smaller post-reform markup Changes in Premiums -$621.4 because of removal of adverse selection -$522 because of smaller post-reform markup Changes in Welfare $335 because of removal of adverse selection (6.4% of insurer medical expenditures) 95% CI [$144, $497] $107 because of smaller post-reform markup Total Net Welfare gain: $71,020,000 per year

42 Optimal Mandate Our estimates suggest the universal coverage is optimal The penalty must be suciently large such that every consumer buys health insurance The minimal tax penalty is $2,190 Larger than Massachusetts penalty but relatively closer to national penalty under the ACA, which can equal the maximum of 2.5% of household income and $2,085

43 Conclusion 1 Develop a simple model of the individual insurance market in MA Model allows us to analyze the impact of an individual mandate on adverse selection Use the model to derive sucient statistics formulas, which allow us to express the welfare eects and the optimal penalty in terms of a small number of empirical moments Use Massachusetts reform to identify these moments and to quantify the welfare eects/ optimal penalty Findings: Welfare gains through the reduction of adverse election: $335 per person ($71 million) Welfare gains because of smaller post-reform markups: $107 per person ($22.7 million) Optimal tax penalty of $2,190 induces universal coverage

44 Conclusion 2 Predictions for the ACA Suggests individual mandate the focus of the SCOTUS challenge is fundamental to mandate-based reform at the national level Suggests that individual mandate enhances welfare due to reductions in adverse selection May depend considerably on existing state-level insurance regulation and how states implement exchanges

45 The Role of the Penalty Tax Penalty Full Welfare Eect Net Welfare Eect Baseline: [162,646] [144,497] [232,735] [123,385] [224,724] [133,415] [216,713] [140,439] [207,701] [144,456] [199,691] [146,471] [190,679] [146,480] [180,669] [146,489] [171,657] [146,494] [152,635] [140,498] GI:

46 Change in the Generosity of Plan Design 1 Measure HMO enrollment via name of insurer. Controlling for HMO enrollment, we nd even larger welfare estimates 2 Evidence from the regulator Minimum Median Maximum Minimum Median Maximum Actuarial Value Deductible $2000 $2000 $2000 $2000 $2000 None Coinsurance N/A N/A N/A N/A N/A N/A PCP Oce Visit $35 $25 $25 $35 $25 $25 SPC Oce Visit $50 $25 $25 $50 $25 $25 Inpatient Copay Deductible Deductible $500 Deductible $500 $800 Outpatient Surgery Copay Deductible Deductible $250 Deductible $250 $250 Emergency Room Copay $200 $100 $75 $200 $75 $100 Pharmacy Deductible N/A None None $250 None None Retail Generic N/A $10 $10 $20 $10 $15 Retail Preferred N/A $50 $30 $50 $30 $30 Retail Non-Preferred N/A $100 $60 $75 $60 $50

47 Welfare Eects using Estimates from the Literature 1 Use premium estimates from Graves and Gruber (2012) 2 Use demand elasticities from Ericson and Starc (2012) 3 Use average cost estimates from Hackmann, Kolstad, and Kowalski (2012) Full Welfare Eect Net Welfare Eect Baseline Estimates Graves and Gruber (2012) Ericson and Starc (2012): Premiums Ericson and Starc (2012): Coverage Hackmann, Kolstad, and Kowalski (2012)

48 Tax Penalties back Above 300% FPL Income and Age % FPL % FPL % FPL Age Age 27+ Tax penalty $19 per month $38 per month $58 per month $83 per month $105 per month $228 per year $456 per year $696 per year $996 per year $1260 per year

49 Impact of Reform on Coverage: NHIS back Group Market Individual Market Year Year Non MA MA Non MA MA