empirical evidence Flashcards
Theme 1 - Relationship between education and adult mortality in US
- Lleras-Muney (2005)
○ 1915-1939 - over 30 states legalised increase in schooling
○ Compared mortality rates before and after schooling laws
○ Additional year of education in 1960 increased life expectancy by 1.7 years at 35 years old - suggests large causal effect
○ Application of study limited to US and similar economies at the time
○ Education level in US low initially at the time before introduction of the schooling laws
○ Unclear how study would apply to countries with high initial education levels
Limited marginal effects of an increase in education on health for those with high education levels compared to low - prioritises importance of education in developing countries
Theme 1 - Relationship between compulsory education and mortality - France
- Albouy and Lequein (2009)
○ Relationship between compulsory education and mortality - estimates causal effects
○ French policy reforms - raised minimum compulsory school leaving age - natural experiment
§ Raise 13-14 for those born in 1923
§ Raise 14-16 for those born in 1953
○ Probability of surviving to age 50 increases with increasing education
○ Study did not find causal impact of education on survival rates
○ Focuses on mortality - doesn’t take into account other returns to health
○ Health returns to schooling more prominent at early ages
○ Returns to health for extra year of schooling at this age low compared to lower ages - not really learning much about health at this age - that is more set at early ages
Ambiguous relationship between education and health
Theme 1 - Differences in health behaviours by education
o Cutler and Lleras-Muney (2010)
o Number of US sources for education and health behaviours
o Focused on white individuals that have completed education, to remove endogenous variables
o MECHANISMS
o Material Resources - allows the purchase of goods that improve health and increase steady-state consumption to raise utility of living to an older age - accounts for 20% of education gradient in behaviours
o Knowledge - direct knowledge (learning at school etc) and information processing (educated better able to use complex technologies/treatments) - accounts for 5-30% of education gradient in behaviours
o Personality traits - self-esteem, self control etc - little impact on education gradient in behaviours
o Social integration - social ties, relationships etc - slight account of education gradient in behaviours
Theme 1 - economic causes of obesity
o Cawley (2015)
o Obesity and income
o Increase in income can either increase or decrease weight
- Weight gain if food is a normal good
- Weight loss if good health is a normal good
Theme 1 - Taxation for SSBs - UK
○ Soft Drinks Industry Levy
○ To encourage the soft drinks industry to improve healthiness of their drinks
○ 24p per litre for drinks over 8g sugar per 100ml,18p per litre for drink 5-8g sugar per 100ml, No charge for those with less than 5g
○ Observations between 2015-19 - tax imposed in April 2016
○ No. of levy eligible drinks fell from 49% to 15% from 2015 to 2019
○ Large impact on amount of sugar in drinks
○ Little change in product size or no. of products available
○ Results not weighted by sales of soft drinks - so changes in sugar consumption habits cannot be estimated
Largely descriptive analysis - difficult to determine a causal relationship from the design
Theme 1 - Taxation for SSBs - US
o Exploits state level changes in taxation on SSBs across time
o Allowed better identification of causal effects
o No effect of taxes on probability of consuming SSBs
o Youths - increase in tax decreases soft drink consumption
o Adults - increase in tax slightly decreased BMI and obesity
o ISSUE - there exist no large tax increases from which to infer behavioural changes
o Substitution to untaxed items may limit impact of the tax on weight/obesity
Theme 2 - socioeconomics inequalities in waiting times for primary care
○ Martin et al (2020)
§ Measures socioeconomics inequalities in waiting times for primary care
§ Used in presence of compulsory public health insurance and limited co-payments, demand for treatment > supply leading to long waits
§ Waiting times act as a form of non-price rationing helping to contain demand by discouraging some patients to seek treatment
§ Rationale for rationing by WTs is an equity concern - should not depend on ability to pay
§ Waiting times measured as time to get an appointment to see a doctor or nurse, SES measured via income, controls for need and if patient holds private health insurance
§ Data from eleven high-income countries
§ Measured through survey questions
□ Last time you were sick, how quickly could you get an appointment for a doctor?
□ What is your household income compared to the average?
§ Variation across countries in WT
□ Waiting > 2 weeks - Canada 14.4%, UK 0.4%
□ Waiting < 2 days - Canada and Norway lowest at 0.4%, New Zealand highest at 0.77%
§ Results
□ Compared to baseline income (much below average), increased income is negatively related to WTs
□ Majority of effects non-significant
§ Conclusions
□ Evidence that patients with higher incomes tend to wait less for access to primary care for Canada, Germany, Norway, and Sweden
□ Potential mechanisms
® People with higher incomes may live in neighbourhoods with greater availability of primary care
® Patients with higher socio-economic status may exercise more choice, and look for more responsive doctors
® May be more insistent when seeking care, and articulate needs fully
® May be more flexible in their employment allowing attendance at first available appointment
□ Potential limitations
® Small sample sizes for countries
® Controls for health care may be insufficient
® Control for whether individual has private health insurance but unsure if this was used
Not clear if inequalities in WTs occur within practices or across practices
Theme 2 - ex ante inequality of opportunity for biomarkers
○ Ex ante IOp to evaluate biomarkers
○ Used ex ante approach - allows proportion of total health outcome inequality to be attributed to IOp
○ Evidence of considerable influence of social and economic patterning in cardiovascular risk
○ Age and gender combined account for largest share of IOp
§ Individuals education 2nd, parental occupational status 3rd, parental education 4th
○ Issues
§ Circumstances often difficult to disentangle
□ Parental characteristics are circumstances
□ Childhood characteristics difficult - cut off at 18 where lifestyle attributed to efforts, but influenced by circumstances in childhood
Educational attainment often defined as a circumstance
Theme 2 - inequity in health care utilisation
o Van Doorslaer and Masseria (OECD, 2004)
o Data - European Household Community Panel 2000
o Reported Health utilisation over the past 12 months - GP visits, hospital admissions etc
o Need indicators - self reported health, health problems
o RESULTS
o Overall access to GP pro poor
o Overall access to specialist services pro rich
Theme 3 - causal estimates of impact of insurance coverage on medical care consumption
- RAND experiment
○ Causal estimates of the impact of insurance coverage on medical care consumption
○ Should US government provide full health insurance coverage or should private insurers design contracts with tailored cost-sharing mechanisms?
○ Observational data suffer from selection bias - individuals who are already insured are likely to need more medical care, and therefore consume more care, than individuals who choose to remain uninsured
○ Biggest randomised experiment ever implemented in economics - funded by the US department of health, education and welfare
○ Methods
§ Enrolled 5800 individuals from 2000 households in 6 different locations in US - representative of US families with adults under the age of 62
§ Randomised experiment design - households are assigned to different levels of cost-sharing
□ Coinsurance levels - 0% (free plan), 25%, 50%, 95%, free inpatient care and 95% for outpatient
§ Maximum dollar expenditure
§ Followed households for 3-5 years and surveyed their medical expenses and health outcomes
○ Results
§ Demand for medical care decreases with higher rates of co-payments
§ RAND estimate of the price elasticity of demand for care: -0.2
§ In patient care price elasticity: -0.14
§ Low income families do not reduce their medical consumption more than high income families
§ Find better health outcomes for low income families in the ‘free plan’
§ RAND experiment not very well designed to study income effects nor health outcomes
○ Discussion/Limitations
§ Threats to internal validity -
□ Non-random assignment to plans
□ Differential participation in the experiment across treatment arms - unbalanced characteristics pre/post randomisation
□ Differential reporting across treatment arms - free plan 4% under reported, 95% 12% under reported
§ Estimates of elasticity
□ Danger of summarising non-linear coverage by a single price
□ Behavioural response is different with non-linear budget sets
○ Policy implications
§ Price elasticity of RAND still used to justify cost-sharing in public and private health care systems as a response to ex post moral hazard
§ Cost-sharing is also used to limit the global increase in health care budgets, especially in tax funded systems
§ Coinsurance rates are sometimes used by regulators to give patients incentives to go for cheaper/more efficient treatments
Impact of OOP can still be catastrophic for households
Theme 3 - targets and terror reducing waiting times
- Propper et al (2008)
○ England implemented ‘command and control’ strategy to tackle long waiting times for elective treatment
§ Targets for inpatient care - from 18 to 6 months max wait time guarantee
§ Dismissal of managers for poor performance against the targets
§ Rewards with greater autonomy if targets are met
○ Method
§ Natural experiment
§ Target and sanctions introduced in England in 2000s but not in Scotland - Scotland used as a diff in diff design
○ Result
§ Target and terror policy reduced proportion of patients waiting more than 6 months by about 20%
○ Conclusions
§ Policy implemented in England effective to reduce long wait times for elective treatment
○ Limitations
§ Not possible to disentangle respective effects of managerial sanctions, escalating targets, and stronger focus on performance
§ Used published data which is not immune to manipulation
§ Quality difference between sets of data for England and Scotland
§ Reduction in waiting times does not automatically imply increase in welfare - reduced wait times does not necessarily affect mean and median waiting times
Possible adverse effects of targets: shorter waiting times achieved by targeting less needy patients or reducing other activities - no clear evidence of either effects (Bevan and Hood, 2006)
Theme 3 - Estimate causal impact of health insurance on health care utilisation, self reported health, financial strain and overall wellbeing
o Oregon Health Insurance Program
o Medicaid program - public health insurance coverage for low income individuals and households
o Implemented through a lottery system
o Data through administrative data and survey results
- Medical consumption - hospital admissions, outpatient visits etc
- Health outcomes - mortality, self reported health and happiness
- Financial strain - measure of unpaid bills, out-of-pocket expenses
o Randomised controlled design
o RESULTS
- Increase in probability of hospital admissions, prescription drugs, and no of outpatient visits
- Significant decline in financial strain measures - OOPs, borrowing money for medical expenses
- Improvements in all measures of good health
o LIMITATIONS
- Internal validity - winning Medicaid lottery could encourage individuals to access more welfare benefits
- External validity - self selection as individuals need to sign up fot lottery, Oregon population not massively representative of US population - aging population, majority white, worse overall health
Theme 5 - effects of reforms on financial protection
- Thailand Reform in 2001 toward universal health insurance
○ Extend publicly financed coverage
○ Adopt supply side measures to constrain costs and deliver cost-effective care
○ Tax-financed, single payer with a fixed budget
○ Capitation payments for outpatient care, prospective budgets for inpatient care - both aim to constrain costs- Prior to 2001
○ 1991 - 2/3 population without formal insurance - reduced to <30% by 2001
○ Medical welfare scheme - covers poor, disabled, children, elderly etc - 32% of population
○ Voluntary health care scheme - households could purchase health card - supplemented by gov subsidy - 21% of population
○ Employment schemes - generous benefits
§ Public sector - government scheme
§ Private sector - social security scheme
○ Those not covered by insurance in 2001 could purchase care through OOP or claim free/subsidised care through being poor - Post reform in 2001
○ All citizens not covered by employment schemes entitled to universal coverage scheme
○ By 2011 98% of population covered by some form of insurance
○ Small OOP ($0.75) with each contact with health services
○ Overall achieved balance between universal coverage on a modest budget and prevented medical expenditure from largely increasing
○ Supply side reforms - capitation-based budgeting, prospective payments of hospital for inpatient treatment, gatekeeper role for access to specialist care, single public purchaser of care - Empirical approach
○ Diff in diff approach
○ Treatment group - households with no public sector employees or not all private sector employees
○ Control group - households in which head is public sector employee - Results
○ Reform reduced total OOP expenditure
○ Probability of incurring OOP increased
○ Large effects for all care types for positive spending
○ Findings support conclusion that reform greatly reduced exposure to medical expenditure - Words of caution
○ Individual level of insurance status or medical expenditure unavailable, only at household level
○ Diff in diff estimates requires pre-treatment parallel trends assumption for treatment and control groups - not directly testable with data used
Lack of impact on inpatient admissions in rural location suggests reform failed to address geographical inequalities in access
- Prior to 2001
