The Impact of Medicaid Coverage on Mental Health, Why Insurance Makes People Happier in OHIE: by Spending Less or by Spending More?

Basic Information

• Paper ID: 2510.14909
• Title: The Impact of Medicaid Coverage on Mental Health, Why Insurance Makes People Happier in OHIE: by Spending Less or by Spending More?
• Author: Yangyang Li (MIT)
• Classification: econ.GN cs.CY econ.TH q-fin.EC
• Conference: Proceedings of the Global Public Health Conference, Vol. 8, Issue 1, 2025
• Paper Link: https://arxiv.org/abs/2510.14909

Abstract

This study leverages the unique randomized controlled design of the Oregon Health Insurance Experiment (OHIE) to investigate the causal relationship between Medicaid coverage and subjective well-being among low-income adults. Based on follow-up survey data at baseline and 12 months, the research focuses on analyzing how Medicaid coverage affects mental health through changes in medical expenditures: whether by increasing or decreasing medical spending to improve well-being. Using instrumental variable (IV) regression methodology, the analysis examines six subgroups classified by expenditure levels and happiness ratings. Results demonstrate that enrollment in the Oregon Health Plan (OHP) significantly reduces the probability of unhappiness regardless of medical expenditure levels, primarily through alleviating financial burden and improving welfare across different subgroups.

Research Background and Motivation

Problem Definition

The core research question addresses how Medicaid coverage affects mental health and subjective well-being among low-income populations, and whether this impact is achieved through increased medical service utilization (high spending) or reduced financial burden (low spending).

Research Significance

1. Policy Relevance: With the expansion of Medicaid eligibility under the Affordable Care Act, understanding its impact on mental health has important policy implications
2. Causal Inference Requirements: Previous observational studies suffer from selection bias, making it difficult to establish causal relationships
3. Theoretical Mechanism Exploration: There is a need to understand the specific mechanisms through which Medicaid affects subjective well-being

Limitations of Existing Research

1. Endogeneity Issues: In traditional observational studies, insurance coverage and health outcomes are subject to unobserved confounding factors
2. Unclear Mechanisms: Existing research lacks in-depth analysis of the specific pathways through which Medicaid affects well-being
3. Neglected Heterogeneity: Systematic analysis of effect variations across different expenditure levels is lacking

Research Motivation

The OHIE provides a rare randomized controlled trial opportunity, with Medicaid eligibility randomly assigned through a lottery system, offering an ideal natural experiment environment for establishing causal relationships.

Core Contributions

1. Innovative Analytical Framework: First systematic analysis of Medicaid's heterogeneous effects across different combinations of happiness levels and expenditure levels
2. Causal Inference Methodology: Utilizes OHIE's randomized design and applies instrumental variable methods to address endogeneity issues
3. Mechanism Identification: Clearly identifies that Medicaid primarily improves well-being through alleviating financial burden rather than increasing medical utilization
4. Policy Insights: Provides empirical evidence on the mental health effects of Medicaid expansion policy

Methodological Details

Task Definition

Input: Individual lottery outcome (instrumental variable Z_i), Medicaid coverage status (treatment variable D_i), control variables (X_i) Output: Probability changes across six different happiness-expenditure combinations Constraints: Instrumental variable relevance, exogeneity, and exclusion restriction assumptions

Model Architecture

Core Identification Strategy

Employs two-stage least squares (2SLS), utilizing lottery results as instrumental variables:

First Stage:

D = γ₀ + γ₁Z + X'γₓ + u

Second Stage:

Ỹⱼ = βⱼ,₀ + βⱼ,₁D + X'βⱼ,ₓ + ϵ, j = 1,···,6

Subgroup Definition

Constructs six binary outcome variables Ỹⱼ,ᵢ:

• Ỹ₁,ᵢ = 1(Yᵢ=1, AMᵢ=1): Unhappy and above-median spending
• Ỹ₂,ᵢ = 1(Yᵢ=1, AMᵢ=0): Unhappy and below-median spending
• Ỹ₃,ᵢ = 1(Yᵢ=2, AMᵢ=1): Fairly happy and above-median spending
• Ỹ₄,ᵢ = 1(Yᵢ=2, AMᵢ=0): Fairly happy and below-median spending
• Ỹ₅,ᵢ = 1(Yᵢ=3, AMᵢ=1): Very happy and above-median spending
• Ỹ₆,ᵢ = 1(Yᵢ=3, AMᵢ=0): Very happy and below-median spending

Technical Innovations

1. Multi-dimensional Decomposition: Decomposes happiness effects by expenditure level to identify heterogeneous effects
2. Rigorous Causal Identification: Leverages exogenous variation from random lottery, satisfying all instrumental variable assumptions
3. Probabilistic Interpretation: Coefficients βⱼ,₁ directly interpret as marginal effects of treatment on subgroup probabilities

Model Assumptions

1. Independent and Identically Distributed: {Yᵢ, Dᵢ, Zᵢ, Xᵢ}ᵢ₌₁ᴺ i.i.d.
2. Exogeneity: E(ϵᵢ|Zᵢ) = 0, E(uᵢ|Zᵢ) = 0
3. Relevance: corr(Dᵢ, Zᵢ) > 0
4. Finite Variance: var(ϵᵢ|Zᵢ) < ∞, var(uᵢ|Zᵢ) < ∞

Experimental Setup

Dataset

• Data Source: OHIE public data from NBER
• Sample Size: 89,824 individuals participated in lottery, 35,169 won, approximately 23,000 in final analysis sample
• Time Span: Lottery conducted March-September 2008, follow-up surveys 1-2 years later
• Data Type: Longitudinal panel data including baseline and 12-month surveys

Key Variables

• Instrumental Variable (Z): Lottery win status (binary)
• Treatment Variable (D): OHP enrollment status (binary)
• Outcome Variable (Y): Happiness level (1=unhappy, 2=fairly happy, 3=very happy)
• Expenditure Variable (E): Out-of-pocket medical spending in past 12 months
• Control Variables (X): Age, gender, household size, baseline health status

Data Processing

1. Sample Selection: Retained individuals participating in both baseline and 12-month surveys, deleted 16,517 records
2. Missing Value Treatment: Corrected 406 cases with missing treatment variable
3. Outlier Treatment: Deleted extreme spending values (2.20×10⁷)
4. Variable Construction: Constructed high/low spending indicators based on median expenditure

Experimental Results

Main Results

Key Findings

1. Consistent Reduction in Unhappiness: OHP significantly reduces the probability of unhappiness regardless of spending levels
2. Moderating Role of Expenditure Level:
   • High-spending group: Reduces probability of "fairly happy," no significant effect on "very happy"
   • Low-spending group: Increases probability of both "fairly happy" and "very happy"
3. Financial Burden Alleviation Mechanism: Effects primarily manifest in low-spending groups, supporting the financial burden alleviation hypothesis

Mechanism Analysis

Results suggest Medicaid affects well-being through the following mechanisms:

• Financial Protection: Reduces economic stress from medical expenses
• Psychological Security: Provides peace of mind from medical coverage
• Not Through Increased Medical Utilization: Limited effects in high-spending groups suggest well-being is not achieved through increased medical services

Related Work

OHIE-Related Research

1. Finkelstein et al. (2012): Demonstrates Medicaid increases medical utilization and reduces financial pressure
2. Taubman et al. (2014): Finds emergency department use increases 0.41 visits per person, challenging conventional assumptions
3. Kowalski (2016, 2018): Applies marginal treatment effect methods to analyze heterogeneity

Health Insurance and Mental Health

Existing research primarily focuses on:

• Changes in medical utilization
• Alleviation of financial stress
• Improvements in health outcomes

This study's unique contribution lies in systematically analyzing the relative importance of different mechanism pathways.

Conclusions and Discussion

Main Conclusions

1. Universal Mental Health Improvement: Medicaid coverage significantly reduces the probability of unhappiness
2. Mechanism Primarily Financial Alleviation: Effects are mainly achieved through reducing financial burden rather than increasing medical utilization
3. Heterogeneous Effects Exist: Different spending-level groups benefit to varying degrees and through different pathways

Policy Implications

1. Support for Medicaid Expansion: Provides empirical evidence of mental health improvements
2. Focus on Financial Protection Function: Policy design should emphasize the role of alleviating financial burden
3. Targeted Interventions: Enables design of differentiated insurance products for different groups

Limitations

1. Short-term Effects: Only observes 12-month effects; long-term impacts remain unknown
2. Geographic Specificity: Oregon results may not fully apply to other regions
3. Mechanism Inference: While results support financial alleviation mechanism, mediating variables were not directly measured
4. Sample Attrition: Approximately 28% sample loss at 12 months may introduce selective attrition bias

Future Directions

1. Long-term Follow-up Studies: Observe effects over 2-5 years
2. Mechanism Verification: Directly measure financial stress, medical accessibility, and other mediating variables
3. Heterogeneity Analysis: Further stratify analysis by demographic characteristics, health status, etc.
4. Cost-Benefit Analysis: Evaluate economic value of mental health improvements

In-Depth Evaluation

Strengths

1. Rigorous Research Design: Utilizes the gold standard of randomized controlled trials with high causal inference credibility
2. Innovative Analytical Methods: Six-category regression framework cleverly captures heterogeneous effects
3. Strong Policy Relevance: Directly addresses important policy questions regarding Medicaid expansion
4. Clear Mechanism Identification: Effectively distinguishes different mechanisms through expenditure-level stratification

Weaknesses

1. Sample Representativeness: Low-income Oregon residents may not fully represent national population
2. Measurement Limitations: Three-category happiness classification may be too coarse, losing information
3. Confounding Factors: While randomization reduces bias, unmeasured covariates may remain
4. Interpretation Depth: Lacks in-depth discussion of why high-spending group effects are limited

Impact

1. Academic Contribution: Provides important empirical evidence for health economics field
2. Policy Value: Offers scientific basis for U.S. health insurance policy reform
3. Methodological Insights: Demonstrates how to utilize natural experiments to study complex policy effects

Applicable Scenarios

• Health insurance policy evaluation
• Social security system design
• Low-income welfare policy
• Mental health intervention program assessment

References

1. Finkelstein, A., Taubman, S., Wright, B., et al. (2012). The Oregon Health Insurance Experiment: Evidence from the first year. The Quarterly Journal of Economics, 127(3), 1057-1106.
2. Taubman, S. L., Allen, H. L., Wright, B. J., Baicker, K., & Finkelstein, A. N. (2014). Medicaid increases emergency-department use: Evidence from Oregon's Health Insurance Experiment. Science, 343(6168), 263-268.
3. Kowalski, A. E. (2016). Doing more when you're running late: Applying marginal treatment effect methods to examine treatment effect heterogeneity in experiments. NBER Working Paper No. 22363.

Assessment Summary: This is a high-quality empirical economics research paper that cleverly leverages OHIE's randomized design to answer important policy questions. The research methodology is rigorous and results have significant policy implications, though there remains room for improvement in interpretation depth and analysis of long-term effects.