Health Expenditures, Longevity and Growth IX European Conference of the Fondazione RODOLFO DE BENEDETTI “Health, Ageing and Productivity” Limone sul Garda, 26 May, 2007 Brigitte Dormont, Joaquim Oliveira Martins, Florian Pelgrin and Marc Surcke 1 Outline of the presentation 1. From Ageing to Longevity. Health ageing offers a potential to translate longevity into active life 2. Determinants of Health spending: ageing & technological progress. Health spending and health outcomes. Optimal health spending 3. Determinants of Health spending: income growth. Is health a luxury good? 4. Projections of total (public+private) health expenditures 2005-2050 5. Health, productivity & growth. Do health status and health spending affect growth? R&D, innovation and global competition for the “health market” 2 1. From Ageing to Longevity: Health ageing offers a potential to translate longevity into active life 3 A major shift in population structure (shares by age group in % total population) EU15 United States 12.0 12.0 ? 10.0 10.0 2000 8.0 8.0 2050 6.0 2000 6.0 4.0 4.0 2.0 2.0 2050 Working age population Working age population 59 10 14 15 19 20 24 25 29 30 34 35 39 40 44 45 49 50 54 55 59 60 64 65 69 70 74 75 79 80 84 85 89 90 94 95 + 04 59 10 14 15 19 20 24 25 29 30 34 35 39 40 44 45 49 50 54 55 59 60 64 65 69 70 74 75 79 80 84 85 89 90 94 95 + 0.0 Japan 12.0 10.0 8.0 2000 6.0 2050 4.0 2.0 Working age population 59 10 14 15 19 20 24 25 29 30 34 35 39 40 44 45 49 50 54 55 59 60 64 65 69 70 74 75 79 80 84 85 89 90 94 95 + 0.0 04 04 0.0 4 Are we underestimating longevity gains? (A) average gains 1960-2000 United States Europe Austria Belgium Czech Republic Denmark 1.7 (B) projected gains 2000-20501 years/decade 1.4 2.4 1.8 1.1 1.1 1.4 1.6 1.3 1.1 Finland France Germany Greece 2.2 2.2 2.0 2.1 1.5 1.8 1.2 0.8 Hungary Ireland Italy Luxembourg 0.9 1.7 2.4 2.2 1.6 0.9 1.8 1.1 Netherlands Poland Portugal Slovak Republic 1.1 1.5 3.1 0.7 0.5 2.0 1.1 1.5 Spain Sweden United Kingdom 2.3 1.7 1.8 0.8 0.9 1.6 EU15 average 2.0 1.2 Japan Memo item: OECD average 3.4 0.8 2.2 1.2 Source: National projections 5 Impact of indexing US working-age population on longevity gains 15-29 50-64 Total 30-49 Additional WAP With longevity indexation 250.0 200.0 150.0 100.0 50.0 2050 2045 2040 2035 2030 2025 2020 2015 2010 2005 2000 1995 1990 1985 1980 1975 0.0 1970 Millions 300.0 6 …and on EU-15 working-age population? 15-29 50-64 Total 300.0 30-49 Additional WAP With longevity indexation 250.0 200.0 150.0 100.0 50.0 2050 2045 2040 2035 2030 2025 2020 2015 2010 2005 2000 1995 1990 1985 1980 1975 0.0 1970 Millions 350.0 7 Impact of longevity indexing on US dependency ratios (65+/15-64) % 50.0 United States 45.0 40.0 With indexation 35.0 Labour force 30.0 25.0 20.0 With indexation Working age population 15.0 2050 2045 2040 2035 2030 2025 2020 2015 2010 2005 2000 10.0 8 …and on EU-15 old-age dependency ratio? 80.0 EU15 70.0 With indexation 60.0 50.0 Labour force 40.0 With indexation 30.0 Working-age Population 2050 2045 2040 2035 2030 2025 2020 2015 2010 2005 20.0 2000 % 9 Indexing the old-age threshold in line with longevity gains would only contribute to solve the ageing problem if aged workers… (1) Remain in good health (“Healthy ageing”) (2) Participate in the labour force and are employed (3) Pension systems are reformed in order to remove incentives for early retirement 10 Road-map of the next sections s5 R&D/Innovation Welfare Health spending/ Investment s2 Technological progress s2 Health status Longevity s2 s5 GDP s4 Income elasticity s3 11 2. Determinants of Health spending: -Ageing & technological progress -Health spending and health outcomes -Optimal health spending 12 2.1 The main driver of health expenditure growth: changes in practices Why ageing impacts health expenditures 28 26 €uros 24 22 20 18 16 14 4000 3500 3000 2500 2000 1500 1000 500 0 2000 0 05 010 015 020 025 030 035 040 045 050 0 2 2 2 2 2 2 2 2 2 2 Population ageing France, 2005-2050 10 20 30 40 50 60 70 80 Age group Health expenditure Per capita & age group, France13 Profile drift between 1992 and 2000 The main part of the story: 4000 3500 €uros 3000 Non demographic effects 2500 1992 2000 2000 1500 1000 500 0 0 10 20 30 40 50 Age group 60 70 80 14 The role of the proximity of death The idea of a boom in health expenditures linked to population ageing is not supported by macroeconometric estimations A non significant influence of age on health expenditures is found (Getzen, 1992; Gerdtham et al.,1992,1998, etc.) Possible explanation: high cost of dying. The correlation between age and health expenditures might be spurious due to the fact that the probability of dying increases with age Once proximity to death is controlled for, age would not influence health expenditures Micro-econometric evidence by Zweifel et al., Seshamani & Gray, etc. 15 Yang et al. (2003): Health expenditures and proximity to death 16 Health expenditures by age group : decedents versus survivors For survivors, the expenditure profile is increasing with age 17 The role of time to death: current consensus (i) Both age and time to death have an influence on health expenditures. (ii) Health expenditure predictions have to include time to death in their modelisation in order to be relevant. This last point is now widely accepted. On US data, Stearns and Norton (2004) show that omitting time to death leads to an overstatement of 15 % for health expenditures, when using projected life tables for 2020. 18 The predominant impact of changes in medical practices Retrospective analysis for France 1992-2000 (DormontGrignon-Huber, 2006) Sample of 3,441 and 5,003 French individuals Micro-simulation methods to evaluate the components of the upward drift in the age profile of health expenditures – Role of changes in morbidity at a given age – Role of changes in practices for given levels of morbidity and age 19 Micro-simulation results (Pharmaceuticals, unconditional consumption) 2000 Changes in morbidity Changes in practices 1992 20 Retrospective decomposition of changes in expenditures (Pharmaceuticals, France 1992-2000) Variation 1992-2000 (%) 67.27 Total demographic change 7.63 -part of structural change 4.61 -part of growing size of the population 3.02 Changes in morbidity -9.24 Changes in practices for a given morbidity 52.24 21 Main results Ageing explains a small part of the rise in health expenditures Changes in practices are the most important driver Evidence of health improvements which induce savings These savings are large enough to offset the increase in costs due to ageing 22 2.2 Innovation and product diffusion in health care The research leading to innovation does not necessarily take place in biomedical sector : lasers, ultrasounds, magnetic resonance spectroscopy, computer, nanotechnology. (Gelijns & Rosenberg, 1994) Two mechanisms : substitution (gain in efficiency) and extension (increasing use of the new technology). – Growth in treatment costs results entirely from diffusion of innovative procedures (Cutler & McClellan, 1996) – Example: treatment of heart attack with bypass surgery and angioplasty. – Other examples: cataract surgery, hip replacement, knee replacement, etc. The orientation of technological progress is not neutral: certain type of innovations will be favoured, depending on the design of the health insurance and on the payment systems implemented by the payers (Weisbrod, 1991) 23 Are medical innovations worth the additional costs? What is the impact of health care on longevity and health? Is the value of the gains in longevity and health larger than the additional costs? 24 The impact of health care on longevity and health Robert Fogel (2003) on 45,000 US veterans: average age of onset of chronic conditions increased by 10 years, while life expectancy increased by 6.6 years. Murphy & Topel: gain in life expectancy in the US: +9 years between 1950 and 2000, of which – + 3.7 years for reduced mortality in heart disease – + 1 year for reduced mortality due to stroke Cutler et al. (2006): between 1984 and 1999 improved medical care for CVD in the US explains – 70 % mortality reduction – 50 % reduction in disability caused by CVD Progress in hip replacement and other surgeries explains decline in disability due to musculoskeletal problems (Cutler, 2003) There is empirical evidence, at least for some conditions, that a quality adjusted price index would not rise but decrease over time 25 Three possible scenarios for future changes in morbidity at a given age 26 2.3 The value of health and the optimal allocation of resources to health expenditures It is important to take into account the value of health for two reasons: to improve the measure of economic growth and welfare public expenditures account for a large share of health expenditures efficient decisions need an appropriate valuation of: – health improvements linked to expenditures – collective preferences for better health and additional years of life. 27 Using the value of life to assess the gains in welfare due to health care The value of a statistical life (VSL) is inferred from risk premiums in the job market or by analysing the markets prices for products that reduce the probability of death from $ 2 millions to 9 millions (Viscusi & Aldy 2003) Value of a year of life : $100,000 (Cutler, 2004) VSL can be used to evaluate the return on new technologies in health care: positive for treatment of heart attack ($70,000/$10,000), depression ($6,000/$1,000), cataract surgery ($95,000/$3,000) VSL can also be used to evaluate global improvements in health. Murphy & Topel (JHE, 2006, Kenneth J. Arrow Award for best paper in health economics published in 2006) assess the value of gains in longevity due to health expenditures . The results is striking: for the US between 1970 and 2000, gains in life expectancy added to wealth a gain equal to about 50 % of the GDP each year. Subtracting the costs due to rising medical expenditures lead to a 28 return equal to 32 % GDP. Assessing the optimal allocation of resources to health expenditures Hall & Jones (2007): the optimal allocation of resources maximizes the expected lifetime utility subject to the budget constraint and the health production function. Budget constraint: the income can be spent on consumption or health Theoretical prediction: the optimal share of income devoted to health care s increases if the value of one year of life rises faster than income. This condition is fulfilled for preferences characterised by a specification of the utility function, with a key parameter γ >1 . A large empirical literature suggests that γ =2. Thus, the rising share of health expenditures is likely to fit collective preferences 29 Simulations: optimal health share increases (Hall & Jones) For γ=1.01 the marginal utility of consumption falls more slowly than the diminishing returns in the reduction of health 30 Summing-up Technological progress, instead of ageing, is the main driver of health expenditure growth. Two mechanisms are involved in technological progress in health care, substitution and extension. The growth in health expenditures is entirely explained by the extension effect: more goods are available and consumed. The diffusion of technologies has led to additional costs but also to more value in terms of longevity and better health it has probably contributed to an increase in welfare. Evaluating the level of health expenditures that maximizes social welfare, one finds that social preferences appear to be in favour of a continuous increase in the share of income devoted to health. Maximizing social welfare requires the development of institutions consistent with the predicted increase in health spending. 31 3. Determinants of Health spending: -Income growth -Is health a luxury good? 32 Is health care a luxury or necessity? Is health care a luxury or a necessity? (Getzen, 2000). The answer depends on the level of analysis: health is a necessity at the individual level and a luxury at the aggregate level Omitted variables typically lead to an overestimation of the income elasticity (Dreger and Reimers(2005), AHEAD, 2006) When additional variables are added (age, time trends) the income elasticity is close or below one 33 Empirical evidence on the income elasticity Individual (micro) Income elasticity Insured Newhouse and Phelps (1976) Hahn and Lefkowitz (1992) ≤0.1 ≤0 less insured/uninsured Falk et al (1933) 0.7 Andersen and Benham (1970) - dental 1.2 AHCPR (1997) - dental 1.1 Regions (intermediate) Fuchs and Kramer (1972) – 33 states, 1966 0.9 Di Matteo and Di Matteo (1998) – 10 Canadian provinces, 1965-91 0.8 Freeman (2003) – US states, 1966-98 0.8 Nations (macro) Newhouse (1977) – 13 countries, 1972 1.3 Getzen (1990) – US, 1966-87 1.6 Schieber (1990) – seven countries, 1960-87 1.2 Gerdtham and Löthgren (2000, 2002) - 25 OECD countries, 1960-97 Dreger and Reimers (2005) – 21 OECD countries Co-integrated Unitary elasticity not rejected 34 Econometric estimation issues Time-series, cross-section or panel analysis? – Evidence is now based on time-series and panel data – Omitted variables, endogeneity, heterogeneity? – Unit root tests and co-integration tests: GDP and Health care expenditure are characterised by unit-roots and are cointegrated. – Cross-sectional dependence (countries are not independent) – Convergence of health expenditures across countries Existence of a third factor? – Co-integration results can be driven by the existence of one or more common factors (technology, population, ...). As seen in section 2, technology is a main driver of health expenditures, but how to capture such an effect? 35 A simple econometric test Dependant variable: log of health expenditures per capita Model I Model II Log GDP per capita 1.58*** 0.937*** -- 0.017*** Time trend NB: 30 OECD countries, for the period 1970-2002. Including one-way fixedeffects. On average, the share of Health expenditures to GDP tends to grow at around 1.7% per year 36 Econometric approach We provide an extensive empirical test: – By decomposing health expenditures (private, public and total) – Use of different country groupings – Include time trends, age structure and some institutional variables – Test for different specifications: pooled, one-way, twoway fixed effects, and random-weight estimators A unitary income elasticity seems the most reasonable assumption to project health expenditures. But this is not small! This implies that the increase in the share of health to GDP is due other factors 37 4. Projections of total (public & private) health expenditures 2005-2050 38 The projection framework is based on health care public expenditure profiles by age-groups % of GDP per capita (normalised GDP p.c. 1999) 25.0 Austria Belgium Denmark 20.0 Finland France Germany Greece Ireland Italy Luxembourg 15.0 Netherlands Portugal Spain Sweden UK Australia United States 10.0 5.0 Age groups 95 + 85 -8 9 90 -9 4 80 -8 4 75 -7 9 70 -7 4 60 -6 4 65 -6 9 55 -5 9 50 -5 4 45 -4 9 35 -3 9 40 -4 4 30 -3 4 25 -2 9 20 -2 4 10 -1 4 15 -1 9 59 04 0.0 39 Source: ENPRI-AGIR and OECD Public vs. Private Health expenditure profiles in the US 7000.0 HE per capita excluding LTC public 6000.0 HE per capita excluding LTC private 5000.0 4000.0 3000.0 2000.0 1000.0 0.0 2 7 12 17 22 27 32 37 42 47 52 57 62 67 72 77 82 87 92 97 Age groups 40 The drivers of expenditure The pure demographic effect : constant expenditure profiles and applied to the change in demographic structures… but this implicitly assumes an “expansion of morbidity” when longevity increases The pure demographic effect has to be adjusted for: – The possibility for different health status [Grunenberg(1977); Fries(1980); Manton(1982)], including a dynamic equilibrium between good health and longevity ("Healthy ageing“) – Which is coherent with the hypothesis that major health costs are concentrated in the proximity to death [eg. Batjlan and Lagergren, 2004] Project expenditures for survivors and non-survivors Non-demographic drivers are the most important 41 Demographic drivers illustrated (1) Pure ageing effect Health expenditure per capita Average in 2050 Average in 2000 Pure demographic effect Young Old Age groups (2) Ageing effect adjusted for death-related costs and healthy longevity Health expenditure per capita Young 42 Old Age groups Non-demographic drivers push expenditure curves up (3) Non-ageing drivers Health expenditure per capita Income + technology residual Non-demographic effects Young Old Age groups 43 Additional exogenous assumptions National population projections (N) [cf. Oliveira Martins et al. (2005)] Labour force projections (L/N) [Burniaux et al. (2003)] Labour productivity (Y/L) growth is assumed to converge linearly from the initial rate (1995-2003) to 1.75% per year by 2030 in all countries, except former transition countries and Mexico where it converges only by 2050. Projected GDP per capita: Y/N = Y/L x L/N The projections allow for a certain convergence of expenditures across-countries 44 Several projection scenarios 2005-2050 (in % of GDP) Scenario I Scenario II η=1 η=1.5 residual=1% p.a. residual=1% p.a. Healthy ageing Healthy ageing Scenario III Scenario IV η=1 residual=2% p.a. Healthy ageing η=1 residual=1% declining to 0 by 2050 Expansion of morbidity (Level 2005) US (14%) 19% 23% 26% 18% EU-15 (8%) 13% 17% 20% 11% Healthy ageing: 1 year gain in life expectancy = 1 year in good health 45 Decomposition of the expenditure change 2005-2050 for EU-15 (in % GDP) Scenario I Scenario II Scenario III Scenario IV η=1 residual=1% p.a. Healthy ageing η=1.5 residual=1% p.a. Healthy ageing η=1 residual=2% p.a. Healthy ageing η=1 declining residual Expansion of morbidity Deathrelated costs 0.2 0.2 0.2 0.2 Pure age effect 1.5 1.5 1.5 1.5 Adj. healthy ageing -0.7 -0.7 -0.7 -- Income effect -- 2.5 -- -- Tech. residual 4.4 4.4 11.4 1.9 46 5. Health, productivity & growth: Do health status and health spending affect growth? R&D, innovation and global competition for the “health market” 47 Health and the economy: main channels Labor productivity: healthier individuals could reasonably be expected to produce more per hour worked Labor supply: Good health increases the number of days available for either work or leisure; Health may influence labour supply (wages, preferences and expected life horizon, but ambiguous effect which depends on substitution and income effects) Education: better health contributes to more educated and productive people; longevity encourage people to invest in education Savings and Investment: health affects savings behavior and willingness to undertake investment R&D and Innovation: Good health enhances creativity and demand for new health goods & services. 48 Empirical evidence Positive impact for developing countries and world level; when measured as life expectancy or adult mortality, health is among very few robust predictors of subsequent economic growth (Levine and Renelt, 1992; Sala-I-Martin, 2004) But mixed evidence for OECD countries (e.g. Rivera and Currais (1999) vs. Knowles and Owen (1995, 1997) regarding life expectancy in OECD countries) 49 Possible explanations Lack of good measure of health status A non-linear relationship (diminishing returns to health) Pension systems and labour markets favoured early retirement, thus the potential effect of better health on participation did not materialise Efforts to increase life expectancy at older ages may have a negative impact on growth. The resources devoted to health care are at the expense of other factors (Aisa & Pueyo, 2005, 2006) An increase of health status is likely to have only a level effect on total productivity, with little impact on labour productivity growth. Assuming contrasted individual ageproductivity profiles have little impact at the macro level. 50 Health and a growth strategy for the EU While EU is doing better in longevity and health status, this potential resources have been wasted in low participation and early retirement of older workers Increasing share of health expenditures to GDP is mainly driven by technological progress. Preferences for longer lives are driving up the optimal share of health spending. Current institutions are not suited to cope with this challenge. There is a large market out there, but EU is lagging in terms of R&D and innovation. This is due to differences in regulation and market structure requiring appropriate product market reforms There strong connections and complementarities across health, labour market, pension reforms, etc. A broad51 reform strategy is needed Thank you ! 52