World Health Organization (WHO) · Publications

Economic consequences of noncommunicable diseases and injuries in the Russian Federation

January 1, 2007 World Health Organization

The full text is hosted by the publishing organisation. lawenc.com indexes the metadata and links to the official source.

Full text

Marc Suhrcke Lorenzo Rocco Martin McKee Stefano Mazzuco Dieter Urban Alfred Steinherr

on Health Systems and Policies

European

Economic Consequences of Noncommunicable Diseases and Injuries in the Russian Federation

The European Observatory on Health Systems and Policies supports and promotes evidence- based health policy-making through comprehensive and rigorous analysis of health systems in Europe. It brings together a wide range of policy-makers, academics and practitioners to analyse trends in health reform, drawing on experience from across Europe to illuminate policy issues.

The European Observatory on Health Systems and Policies is a partnership between the World Health Organization Regional Office for Europe, the Governments of Belgium, Finland, Norway, Slovenia, Spain and Sweden, the Veneto Region of Italy, the European Investment Bank, the Open Society Institute, the World Bank, the London School of Economics and Political Science and the London School of Hygiene & Tropical Medicine.

Economic Consequences of Noncommunicable Diseases and Injuries in the Russian Federation

Marc Suhrcke, Lorenzo Rocco, Martin McKee, Stefano Mazzuco, Dieter Urban and Alfred Steinherr

Keywords: CHRONIC DISEASE – economics WOUNDS AND INJURIES – economics COST OF ILLNESS RUSSIAN FEDERATION

The views expressed by authors or editors do not necessarily represent the decisions or the stated policies of the European Observatory on Health Systems and Policies or any of its partners.

The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the European Observatory on Health Systems and Policies or any of its partners concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Where the designation “country or area” appears in the headings of tables, it covers countries, territories, cities, or areas. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement.

The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by the European Observatory on Health Systems and Policies in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters.

The European Observatory on Health Systems and Policies does not warrant that the information contained in this publication is complete and correct and shall not be liable for any damages incurred as a result of its use.

ISBN 9 789289 021906

Printed in the United Kingdom by The Cromwell Press, Trowbridge, Wilts.

Please address requests about the publication to: Publications, WHO Regional Office for Europe, Scherfigsvej 8, DK-2100 Copenhagen Ø, Denmark

Alternatively, complete an online request form for documentation, health information, or for permission to quote or translate, on the Regional Office web site (http://www.euro.who.int/PubRequest)

Contents

List of tables, figures and boxes vi About the authors ix Acknowledgements xi Executive summary xiii

1 Introduction 1

2 Conceptual framework 5

3 Adult ill-health in the Russian Federation 7

4 Empirical evidence on the economic impact of health in the 11 Russian Federation

5 Further action 45

6 Conclusions 49

Appendix: Description of micro datasets used 51 Details of calculations on the costs of absenteeism 53 Detailed results on the impact of health on labour supply and productivity 54

References 65

List of tables, figures and boxes

Tables

Table 3.1 Life expectancy and adult mortality in selected countries 8

Table 3.2 Life expectancy and healthy life expectancy in the Russian 10 Federation

Table 4.1 Costs of absenteeism due to illness in the Russian Federation 14

Table 4.2 Results of Cox regression model on age to retirement 24

Table 4.3 Random effects logit regression results 25

Table 4.4 Panel probit results on alcohol as a determinant of being fired 28

Table 4.5 Regression results on the effect of a household member’s 29 death on depression

Table 4.6 Regression results on alcohol consumption in response to a 29 household member’s death

Table 4.7 Results from difference-in-differences estimator combined with 31 propensity score technique: effect of adverse health on total income for different periods

Table 4.8 Cause-specific adult death rates in the Russian Federation and 35 EU Member States before May 2004 (age 15–64, per 100 000)

Table 4.9 Economic benefit estimation for most optimistic scenario 36

Table 4.10 Economic benefit estimation for intermediate scenario 37

Table 4.11 Welfare benefits of most optimistic and intermediate scenarios 40

Table 4.12 Growth regression results 43

Table A.1 Calculation for costs of absenteeism 53

Table A.2 Independent variables used in the regression analysis 55 (RLMS data)

Table A.3 OLS – dependent variable: log hourly wage rate (2000 prices) 56

Table A.4 OLS – dependent variable: log weekly hours 57

Table A.5 RLMS IV regression results – dependent variables: log deflated 58 wage rate (2000 prices) and log weekly worked hours (using self-reported health)

Table A.6 RLMS IV regression results – dependent variables: log deflated 59 wage rate (2000 prices) and log weekly worked hours (using work-days missed owing to illness)

Table A.7 NOBUS IV regression results – dependent variable: log monthly 60 wage rate

Table A.8 NOBUS IV regression results – dependent variable: log weekly 60 worked hours

Table A.9 PANEL – dependent variable: log deflated wage rate 61 (2000 prices): males

Table A.10 PANEL – dependent variable: log weekly worked hours: males 62

Table A.11 PANEL – dependent variable: log deflated wage rate 63 (2000 prices): females

Table A.12 PANEL – dependent variable: log weekly worked hours: females 64

Figures

Figure 2.1 From health to wealth (and back) 5

Figure 3.1 Male adult mortality and gross national income (GNI) per capita 8 in 2000

Figure 3.2 Cardiovascular mortality rates in the Russian Federation as a 9 percentage of those of Sweden

Figure 3.3 Injury mortality rates in the Russian Federation as a percentage 9 of those of Sweden

Figure 4.1 Annual average days of absence due to illness per employee in 14 the Russian Federation (2000–2003) and EU Member States before May 2004 (2000)

Figure 4.2 Probability of remaining in the workforce with and without chronic 23 illness, by age, based on Cox regression model

Figure 4.3 Average predicted probability of retiring in the subsequent period, 26 hypothetical male at varying income levels: based on panel logit model

Figure 4.4 Standardized death rates due to CVD and external causes in the 34 Russian Federation (age 0–64, per 100 000)

Figure 4.5 Three scenarios for Russian adult NCD and injury mortality rates 34 (2002–2025) and those of the EU Member States before May 2004 (2001) (age 15–64, per 1000)

Figure 4.6 GDP per capita (in US$ PPP) forecasts in the three scenarios 42

Figure 4.7 GDP per capita (in US$ PPP) forecasts based on OLS and FE 44 regression

Boxes

Box 4.1 Technical details and results of the impact of ill-health on labour 16 supply and productivity

List of tables, figures and boxes vii

Box 4.2 Cox regression: technical details and results 23

Box 4.3 Technical details and results of panel probit model on the 27 probability of being fired

Box 4.4 Technical details and results of household income impact 30

Box 4.5 Technical details and results of economic growth impact 42 estimates

List of tables, figures and boxesviii

Martin McKee is Professor of European Public Health at the London School of Hygiene & Tropical Medicine (LSHTM), where he co-directs the School’s European Centre on Health of Societies in Transition, and he is also a research director at the European Observatory on Health Systems and Policies. His main fields of research include health systems, the determinants of disease in populations, and health policy, all with a focus on eastern Europe and the former Soviet Union.

Stefano Mazzuco, PhD, is Research Assistant in the Department of Statistical Sciences at the University of Padova. He obtained a PhD from the University of Padova in 2003. His main current research interests are demographic economics, with special reference to poverty and family formation and transition to adulthood.

Lorenzo Rocco, PhD, is Assistant Professor of Economics with the University of Padova in Italy. He obtained a PhD from the University of Toulouse I in 2005. His main current fields of research are development economics and health economics.

Alfred Steinherr is Head of the Department of Macro-Analysis and Forecasting at the German Institute for Economic Research (DIW Berlin), Professor of Economics at the Free University of Bolzano, executive in residence and professor at the Sacred Heart University Luxembourg, and with the European Investment Bank, Luxembourg. His main current fields of research include business cycle forecasting, labour economics, and financial markets.

About the authors

Marc Suhrcke, PhD, is an economist with the WHO Regional Office for Europe in Venice, Italy, where he is in charge of the Health and Economic Development workstream. His main current research interests are the economic consequences of health, the economics of prevention and the socio- economic determinants of health.

Dieter Urban, PhD, is Assistant Professor for Economics at Johannes Gutenberg University in Mainz, Germany. He obtained his PhD from Copenhagen Business School and previously held research positions at the London School of Economics and Bocconi University. He teaches panel data econometrics to graduate students. In his research, he undertakes macro- and microeconometric studies in many fields of economics, including health economics. He is also an affiliate of CESifo in Munich.

About the authorsx

Acknowledgements

The work on this report was undertaken in large part as input into the World Bank report Dying too young: addressing premature mortality and ill-health due to noncommunicable diseases and injuries in the Russian Federation, published in 2005.

The World Bank’s support for the contribution of Lorenzo Rocco and for two consultative visits to Moscow by Marc Suhrcke and one by Martin McKee is gratefully acknowledged. We have particularly appreciated the very active support and encouragement of Patricio Marquez (World Bank). Charles Griffin, Cem Mete, Edmundo Murrugara, Willy De Geyndt, Christoph Kurowski, Derek Yach and John Litwack provided very useful and extensive comments on a previous draft. Many thanks go to Elizabeth Goodrich and Nicole Satterley for copy-editing the text. Many of the results presented are the direct output of parallel work coordinated and undertaken by Marc Suhrcke, Lorenzo Rocco, and Martin McKee on a forthcoming report on health and economic development in eastern Europe and central Asia. Dieter Urban, Stefano Mazzuco, and Alfred Steinherr made key contributions to the present report. Financial support for the contribution of the latter three co-authors has been provided by the WHO European Office for Investment for Health and Development in Venice, Italy. Thanks also to Andrea Bertola for support on a number of data issues and Theadora Koller (both WHO Venice Office) for editorial advice. We are also grateful to Giovanna Ceroni from the European Observatory for managing the publication process.

All remaining errors are the sole responsibility of the authors. Views expressed here are exclusively the authors’ and do not necessarily correspond to the official views of their affiliated organizations.

There is increasing evidence of the two-way relationship between health and economic growth. While economic development can lead to improved population health, a more healthy population can also drive economic growth. Similarly, at the level of the individual, while greater wealth contributes to better health, good health is an important determinant of economic productivity. This finding has important policy implications: national and international policy-makers interested in promoting the economic development of a country should seriously consider the role health investment could play in achieving their economic policy goals. Yet little is known about the direct relevance of these recent findings for the transition countries in central and eastern Europe and the Commonwealth of Independent States (CIS) that are facing a very particular health challenge, predominantly comprising noncommunicable diseases (NCD) and injuries. To date, their economic implications have hardly been analysed. This study takes a first step towards analysing the issue. The focus is the Russian Federation, although the findings are also relevant to other transition economies. In particular, we begin to answer two important questions.

• What effect has adult ill-health, in particular NCD and injuries, had on the Russian economy and the economic outcomes of the people living there?

• If the excessive burden of adult ill-health in the Russian Federation were reduced, what economic benefits could result?

The overarching message from our findings is unambiguous: poor adult health negatively affects economic well-being at the individual and household levels in the Russian Federation; and, if effective action were taken, improved health would play an important role in sustaining high economic growth rates.

Executive summary

Our findings relating to the first question are as follows.

• A simple, conservative estimate indicates significant costs of absenteeism due to illness.

• Ill-health appears to have had a significant and sizeable impact on labour productivity in recent years, but less so on labour supply.

• However, the labour supply has been significantly and sizeably affected to the extent that jobholders suffering from chronic illness have retired as a result.

• Severe alcohol consumption significantly increases the probability of losing one’s job.

• The death of a household member affects surviving household members’ welfare and behaviour in at least two ways, i.e. by increasing the probability of depression and of increased alcohol consumption.

• Chronic illness has negatively affected household incomes, particularly during the period 1998–2002.

The second part of this study assesses the macroeconomic benefits that would accrue by reducing NCD and injury mortality rates among adults in the Russian Federation. The main conclusion is that these benefits would be substantial for the Russian economy, irrespective of how they are evaluated. This occurs despite the fact that we assess only the effect of mortality reductions, setting aside morbidity reduction, which would probably attend mortality improvement and almost certainly also be sizeable. Our main findings are set out here.

• The static economic benefit (i.e. valuing a life year gained by one gross domestic product (GDP) per capita) of gradually bringing the Russian Federation’s adult NCD and injury mortality rates down to the most recent rates for European Union (EU) Member States (those belonging to the EU prior to May 2004) by 2025 is estimated to be between 3.6% and 4.8% of the 2002 Russian GDP.

• The broadly defined “welfare” benefits (i.e. using a “value of life” measure) from achieving the rates of the EU Member States (those belonging to the EU prior to May 2004) by 2025 are estimated to be as high as 29% of the 2002 Russian GDP.

• The dynamic benefits (i.e. the effect on economic growth rates) are massive and growing over time. Even if the future returns are discounted to the starting-year value (2002), they represent a multiple of the static GDP effects.

Executive summaryxiv

The third part of the study briefly examines the potential response to the findings obtained, identifying some of the institutional barriers to effective action and setting out some of the policy options.

We have not directly taken into account the costs of different health interventions, the next logical step towards a full economic assessment, but the expected economic benefits would easily exceed any reasonable increase in investments to maintain and promote health, both inside and outside the health system. Another logical step will be to assess the benefits that would accrue from the morbidity reductions expected from those same investments.

These findings have obvious implications for economic and health policy- makers in the Russian Federation as well as for international organizations interested in the country’s social and economic development: investing in the health of the Russian adult population should be seriously considered as one (of several) means by which to achieve economic policy goals. Furthermore, while the analyses were possible in the Russian Federation because of the existence of appropriate data, it is likely that similar findings would be obtained from other economies in transition, given the similarity of their health and economic situations. Hopefully, this report will be a stimulus to other countries in the region to reassess the priorities they place on investment in health as one of the drivers of economic growth.

Executive summary xv

There is increasing evidence of the two-way relationship between health and economic growth. While economic development can lead to improved population health, a more healthy population can also drive economic growth.1 This has important policy implications: national and international policy-makers interested in promoting the economic development of a country should seriously consider the role that health investment could play to further the achievement of their economic policy goals.

Little is known about the relevance of these recent findings for the Russian Federation. Yet it is difficult to believe that the Russian economy and the individuals disproportionately hit by ill-health would not face a severe economic penalty. The Russian Federation is one of very few countries where life expectancy has been decreasing in recent years (McMichael et al. 2004), and the Russian Federation’s health status compares very unfavourably with those of its economic competitors.

Direct evidence that actually measures the impact of poor health on the Russian economy, or, by extension, the gains that might be achieved by reducing avoidable disability and premature death, is scant. One exception is a study (Ladnaia, Pokrovsky and Rühl 2003) that estimates the impact of different scenarios for progression of the HIV/AIDS epidemic on the Russian Federation’s macroeconomic prospects. The authors quantify the prohibitive price that the Russian Federation would pay, in foregone economic growth, if the HIV/AIDS epidemic were left unchecked. Yet while HIV/AIDS is an

Chapter 1

Introduction

1 This case was made cogently in the 2001 report of the Commission on Macroeconomics and Health (CMH 2001) for the developing country context, and more recently, the evidence on the economic benefits of health for high-income countries was assembled in Suhrcke et al. (2005).

extremely serious threat to both the health and economy of the Russian Federation, the predominant share of the current disease and mortality burden involves NCD and injuries. Shkolnikov, McKee and Leon (2001) report that it is not only the historically low level of life expectancy but also the recent reduction in life expectancy that have been driven by mortality from cardiovascular disease (CVD) and injuries. It is also apparent that much of the premature mortality in the Russian Federation occurs disproportionately at ages between 40 and 55, normally a person’s most productive years.

What, then, is the effect of ill-health among adults in the Russian Federation, in particular that due to NCD and injuries, on both the Russian economy and the economic prospects of individual Russians and their families? And what would be the economic benefits of reducing the very high burden of disability and premature mortality among Russian citizens?

This report provides an overview of a series of newly undertaken studies of the economic consequences of health in the Russian Federation, conducted within the framework of a World Bank-led study of the Russian mortality crisis (World Bank 2005). To the best of our knowledge, this is the first comprehensive effort to provide empirical evidence on the economic consequences of adult ill-health in the Russian Federation. We focus on NCDs and injuries, which, according to the World Bank (2005), account for most of the Russian Federation’s ill-health. While there is clearly room for further work on this subject, the message from these analyses is unambiguous: the poor adult health of Russians negatively affects economic well-being at the individual and household levels, and improving adult health can be expected to contribute significantly to sustained economic growth.

In presenting our results on the impact of health on the economy, we are also aware that the relationship between health and economic growth works both ways. It is explicitly not our purpose to argue that the contribution of health to the economy is more important than the contribution of the economy to health. Whether one is more important than the other is debatable and is in any case unnecessary to ask. Here, it is sufficient to show that there is certainly a path from health to the economy. It is this mutually reinforcing relationship between health and the economy that provides a higher return from investing a given amount of resources in both, compared with investing the same amount in either.

Although based on data from the Russian Federation, the findings from these analyses also contribute to our understanding of the economic implications of NCD and injuries in other countries. There is comparatively little research on this subject, in particular in relation to low- and middle-income countries.

Economic impact of health in the Russian Federation2

The relative lack of a convincing economic argument for investing in policies that will combat NCD and injuries may help to explain why these conditions have had so little attention from policy-makers (Yach and Hawkes 2004).

This study is structured as follows. In Chapter 2, we briefly sketch a conceptual framework that highlights some of the channels through which health determines economic outcomes. Chapter 3 presents the key epidemiological facts about adult health in the Russian Federation. Some of these facts already provide highly suggestive evidence of the potential economic importance of adult (ill-)health in the Russian Federation. Chapter 4 presents the core of this study: the empirical evidence on the micro- and macroeconomic impact of health in the Russian Federation. Chapter 5 examines a potential response to the findings obtained, identifying barriers to effective action and setting out some policy options. Chapter 6 presents our convictions derived from the findings presented in the earlier chapters.

Economic impact of health in the Russian Federation 3

Figure 2.1 shows the channels through which health could contribute to an economy and ultimately to economic growth. Four channels are shown, though others may exist: enhanced labour productivity, higher labour supply, higher skills from better education and training, and more savings available for investment in physical and intellectual capital. Figure 2.1 also illustrates that as an economy develops, health improves.

Labour productivity

Healthier individuals could reasonably be expected to produce more per hour worked. First, productivity would be increased directly by enhanced physical and mental activity. Second, more physically and mentally active individuals would make better and more efficient use of technology, machinery, and

Chapter 2

Conceptual framework

Figure 2.1 From health to wealth (and back) Source: Modified from Bloom, Canning and Jamison 2004.

HEALTH

Labour productivity

Labour supply

Education

Saving and investment

ECONOMY

equipment. A healthier labour force could also be expected to be more flexible and adaptable to changes (e.g. in job tasks and the organization of labour), reducing job turnover with its associated costs (Currie and Madrian 1999).

Labour supply

Somewhat counter-intuitively, economic theory predicts a more ambiguous impact of health on labour supply. The ambiguity results from two effects that offset each other. The first effect – the substitution effect – suggests that as lower productivity from poor health leads to reduced wages, workers will respond to the lower returns by working less. Thus, as more leisure is pursued, the labour supply is constricted. The second effect – the income effect – suggests that as poor health leads to reduced wages, workers will work more to recoup lost income, thus expanding the labour supply. The income effect is likely to gain importance if the social benefit system fails to cushion the effect of reduced productivity on lifetime earnings. The net impact of the substitution and income effects thereby ultimately becomes an empirical question (Currie and Madrian 1999).

Education

Human capital theory suggests that more educated individuals are more productive (and obtain higher earnings). Accepting that children with better health and nutrition achieve higher education attainments and suffer less from school absenteeism or dropping out of school early, improved health in early years would contribute to raising future productivity. Moreover, if good health is also linked to higher life expectancy, healthier individuals would have more incentive to invest in education and training, as the rate of depreciation of the gains in skills would be lower (Strauss and Thomas 1998).

Saving and investment

The health of an individual or a population is likely to have an impact not only on the level of income but also on the distribution of income among savings, consumption, and investment. Individuals in good health are likely to have a wider time horizon, so their savings ratio may be higher than that of individuals in poor health. Therefore, a population experiencing a rapid increase in life expectancy may, other things being equal, be expected to have higher savings. This should also result in a higher propensity to invest in physical or intellectual capital (Bloom, Canning and Graham 2003).

Economic impact of health in the Russian Federation6

The Russian Federation is one of only a few countries where life expectancy is falling. However, the situation in the Russian Federation and its ex-Soviet neighbours differs from some other countries where life expectancy is also falling, such as in sub-Saharan Africa, where the declines have been driven by the HIV/AIDS epidemic. In the former, both the recent declines and the current low level of life expectancy were driven largely by increasing mortality among people of working age, with the greatest contribution from NCD and injuries (Shkolnikov et al. 2004; Nolte, McKee and Gilmore 2005). As a consequence, the global development agenda, driven by the pursuit of the Millennium Development Goals (MDGs), may not be perfectly appropriate for the Russian case (and for most other eastern European countries). A recent World Bank report showed how reducing mortality from CVD and injury would have a much greater impact on life expectancy than achieving the health-related MDGs (child and maternal mortality, reductions in HIV/AIDS and tuberculosis (TB)) (Lock et al. 2002; Rechel, Shapo and McKee 2004).

The scale of the challenge is apparent from Table 3.1, which shows that although male life expectancy at birth in the Russian Federation is about two years less than in Brazil or Poland, the probability that a 15-year-old Russian boy will die before he reaches 60 is over 40%, about 16 percentage points higher than in Brazil, double the rate of Turkey, and quadruple that of the United Kingdom.

The fact that a major determinant of a population’s health is its country’s level of economic development may in part explain some of the differences in mortality rates depicted in Table 3.1. However, as Figure 3.1 shows, even if we take income differences into account, Russian male mortality rates are still

Chapter 3

Adult ill-health in the Russian Federation

substantially higher than those of other countries with a similar level of per- capita income. The only countries that are on a still higher trajectory than the Russian Federation are those that have suffered some of the worst HIV/AIDS crises (e.g. Botswana, South Africa, Namibia, Swaziland).

The social consequences of this high toll of avoidable mortality are bound to be significant. The widely held view that NCDs exclusively strike people that have passed retirement age is mistaken. In the Russian Federation the young and the middle-aged are by far more affected than in western Europe. Figure 3.2 illustrates this point by displaying the ratio of mortality in the Russian Federation from CVD in different age groups to that in Sweden. While the death rate is between two and three times higher in older ages, it is a remarkable 12 times higher in the 30–34 age group. A similar, slightly less acute difference is seen for deaths from injuries (Figure 3.3).

Economic impact of health in the Russian Federation8

Table 3.1 Life expectancy and adult mortality in selected countries

Country Life expectancy Probability of dying Probability of dying at birth (years) between ages 15 between ages 15 total (2001) and 60 (% males) and 60 (% females)

(2000 to 2001) (2000 to 2001)

Russian Federation 66 42.4 15.3 Japan 81 9.8 4.4 France 79 13.7 5.7 United States 78 14.1 8.2 Germany 78 12.6 6.0 United Kingdom 77 10.9 6.6 Denmark 77 12.9 8.1 Mexico 73 18.0 10.1 Poland 70 22.8 8.8 Turkey 70 21.8 12.0 Brazil 68 25.9 13.6 Kyrgyzstan 66 33.5 29.9

Source: World Bank 2003.

6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 4.0

4.5

5.0

5.5

6.0

6.5

7.0

M al

e ad

ul t m

or ta

lit y

20 00

(i n

lo g)

GNI per capita PPP 2000 (in log)

Russian Federation

Figure 3.1 Male adult mortality and gross national income (GNI) per capita in 2000 Note: Squares indicate countries in eastern Europe and central Asia. PPP: purchasing power parity Source: World Bank 2004.

The difference between the Russian Federation and western Europe is even greater when their morbidity rates are compared (Table 3.2). An analysis of healthy life expectancy – i.e. life expectancy augmented by a morbidity component – demonstrates the less well-recognized phenomenon of a high level of ill-health among women, in particular those of working age. Indeed, the difference in healthy life expectancy between the Russian Federation and western Europe is even higher than that for life expectancy alone. This confirms that morbidity data contain important information not captured by mortality/life expectancy data. If the Russian health crisis is not merely a health crisis of men, as these findings very strongly suggest, then the economic costs of ill-health are most likely also felt among women.

Economic impact of health in the Russian Federation 9

1400%

1200%

1000%

800%

600%

400%

200%

20 –2

25 –2

30 –3

35 –3

40 –4

45 –4

50 –5

55 –5

60 –6

65 –6

70 –7

75 –7

80 –8

4 85

Figure 3.2 Cardiovascular mortality rates in the Russian Federation as a percentage of those of Sweden Source: WHO Regional Office for Europe 2006.

1400%

1200%

1000%

800%

600%

400%

200%

0% <1

1– 4

5– 9

10 –1

15 –1

20 –2

25 –2

30 –3

35 –3

40 –4

45 –4

50 –5

55 –5

60 –6

65 –6

70 –7

75 –7

80 –8

4 85

Figure 3.3 Injury mortality rates in the Russian Federation as a percentage of those of Sweden Source: WHO Regional Office for Europe 2006.

In sum, this chapter shows that the health challenges facing the Russian Federation affect not only the retired, but also working-age people – and very much so. Moreover, in contrast to what mortality data alone tell us, women’s health has been seriously affected, too. This purely epidemiological evidence alone would suggest that ill-health during middle age has a substantial impact on economic outcomes at the individual and aggregate levels. Chapter 4 examines this issue in depth.

Economic impact of health in the Russian Federation10

Table 3.2 Life expectancy and healthy life expectancy in the Russian Federation

Country/Region At age 20 At age 40 At age 65 LE HLE LE HLE LE HLE

Males Russian Federation 41.9 36.7 22.4 17.3 11.4 6.7 Western Europe 54.5 50.4 31.2 27.6 15.0 12.5

Females Russian Federation 54.2 40.6 31.1 18.5 15.2 5.8 Western Europe 60.2 53.7 36.0 30.3 18.1 14.0

Female–male gap (years) Russian Federation 12.3 3.9 8.7 1.2 3.9 -0.9 Western Europe 5.7 3.3 4.8 2.7 3.1 1.5

Notes: HLE (healthy life expectancy) is calculated by Sullivan’s method (Sullivan 1971); LE: life expectancy. Source: Andreev, McKee and Shkolnikov 2003.

This chapter presents a selection of empirical evidence on various channels through which health has contributed or might contribute to a number of economic outcomes in the Russian Federation. Section 4.1 focuses specifically on the assessment of the economic impact of ill-health in the Russian Federation in recent years. Most of our evidence in this chapter is microeconomic, as this is the level of analysis that most readily allows assessment of the economic impact of ill-health.2 Section 4.2 looks forward by estimating the likely future economic benefits that could be reaped from improving adult health in the Russian Federation in three plausible scenarios.

4.1 What has been the impact of adult (ill-)health on economic outcomes?

After assessing the impact of adult health on economic status in the Russian Federation, our main findings are as follows:

• A simple, conservative estimate indicates significant costs of absenteeism due to illness.

• Ill-health appears to have had a significant and sizeable impact on labour productivity in recent years, but less so on labour supply, at least among jobholders.

Chapter 4

Empirical evidence on the economic impact of

health in the Russian Federation

2 We have not undertaken a macroeconomic assessment of the impact of health on the past macroeconomic performance since the onset of transition in the early 1990s, because we believe that it would be very hard to detect such a causal impact of health in this very transitional period. Nor do we focus on the role of health in determining economic outcomes in the pre-transition period. This has been carried out, for instance, by Davis (2005). He shows that there is much to suggest that the early post-Second World War health achievements made in the former Soviet Union contributed significantly to the comparatively strong economic development in the period up to the early 1970s. However, these health achievements were made in the area of communicable diseases and child and maternal health, not in NCDs and injuries.

• However, the labour supply has been significantly and sizeably affected to the extent that jobholders suffering from chronic illness have retired as a result of such illness.

• Severe alcohol consumption significantly increases the probability of losing one’s job.

• The death of a household member affects surviving household members’ welfare and behaviour in at least two ways, i.e. by increasing the probability of depression and increased alcohol consumption.

• Chronic illness has negatively affected household incomes, particularly during the period 1998–2002.

Since the most visible economic impact of adult ill-health runs via the labour market, we pay most attention to this mechanism (Subsection 4.1.1). Subsequently, we briefly explore the broader impact of chronic illness health on income (Subsection 4.1.2) and the effect of adult mortality on the remaining household members (Subsection 4.1.3).

4.1.1 The impact of health on labour market outcomes

We attempt here to assess various ways in which (ill-)health has had an impact on the labour market in the Russian Federation. It appears intuitively obvious that an individual’s health status has an impact on labour supply – i.e. the number of hours worked and the decision whether to participate in the labour force – as well as on labour productivity – the output produced per unit of time worked (commonly proxied by the hourly or daily wage rate). However, what seems obvious is not always the outcome of scientific reasoning and research. As explained in Chapter 2, economic theory predicts an unequivocally negative impact of ill-health on labour productivity, but an ambiguous one on labour supply.

In what follows we summarize our results on the impact of ill-health on labour supply and productivity in the Russian Federation. In particular, we present estimates of the impact of health status on labour supply and productivity, using two different datasets: the Russian Longitudinal Monitoring Survey and the NOBUS Household Survey (the Appendix has descriptions of the datasets). Moreover, we propose an estimate of the effects of chronic illness on early retirement (as one dimension of labour supply) based on data from the Russian Longitudinal Monitoring Survey (RLMS) and, by slightly changing the perspective, the effects of alcohol on the probability of being fired. This is followed by an attempt to estimate the impact of adult mortality on surviving

Economic impact of health in the Russian Federation12

household members in terms of probability of subsequent alcohol consumption and depression. Finally, we produce a longitudinal analysis of the effect of health problems on income.

4.1.1.1 The cost of work absenteeism due to illness

On average 10 days are lost per employee per year due to illness in the Russian Federation, while the average for Member States belonging to the EU prior to May 2004 is just below 8 days. Work absenteeism due to illness is a widely used, if imperfect, illustration of the effect of illness on the labour supply of individuals. In the 15 Member States belonging to the EU prior to May 2004, for instance, a survey conducted in 2000 found that on average 17% of workers reported having been absent from work at least once in the previous 12 months due to health problems (European Foundation for the Improvement of Living and Working Conditions 2001). These absences represent an average loss of 7.9 working days per worker. Sickness absence incurs the direct cost of the sickness benefits paid to absent employees (when applicable) as well as the indirect cost of lost productivity during the time away from work. In the United Kingdom in 1994, lost productivity due to sickness absence was assessed at over £11 billion (€15.8 billion). In Portugal, 5.5% of all working days in the 2000 largest enterprises were assessed to have been lost in 1993 as a result of illness and accidents. In Belgium, €2.8 billion was paid in 1995 in sickness benefits and benefits for work-related injuries and occupational diseases. In 1993, payments to cover work absence were assessed to be up to €30.6 billion in Germany and €15.8 billion in the Netherlands (€3.9 billion for benefits for sickness absence and €11.9 billion for disability benefits) (see European Foundation for the Improvement of Living and Working Conditions (1997) for data on costs of absenteeism). Figure 4.1 shows the annual average number of work days missed due to illness in the Russian Federation (2000) – calculated using data from the RLMS – compared with the latest available figures for the 15 Member States belonging to the EU prior to May 2004. Although this indicator has a disadvantage in that it reflects both the burden of ill-health and the incentives created by the employment policy environment, it does serve as a useful illustrative example.

The overall cost associated with the reported work days lost due to illness in the Russian Federation varies between 0.55% and 1.37% of GDP, depending on the estimation method (Table 4.1). The monthly absenteeism figures from Figure 4.1 can be converted to a monetary value either by using the average wage rate (resulting in the lower value: column 1) or the GDP per capita (resulting in the higher value: column 3) (details of the calculations are in given in Table A.1 in the Appendix). This is a significant impact, given that the indicator fails

Economic impact of health in the Russian Federation 13

to capture the many other ways in which ill-health has an impact on the labour market. In particular, it does not take into account the effect of the reduction of productivity, nor does it capture the impact on mortality. In a theoretical model, Pauly et al. (2002) examine the magnitude and incidence of costs associated with absenteeism under a range of assumptions (size of the firm, the production function, the nature of the firm’s product, and the competitiveness of the labour market). They conclude that the cost of lost work time can be substantially higher than the wage when perfect substitutes are not available to replace absent workers, when production involves teamwork, or where a penalty is associated with failing to meet an output target.

Economic impact of health in the Russian Federation14

10.8

9.2 9.5

10.9

8.6

10.3 9.6 9.4

0 2000 2001 2002 2003

EU15: 7.9 (M: 6.9, F: 9.0)

Male Female

Figure 4.1 Annual average days of absence due to illness per employee in the Russian Federation (2000–2003) and EU Member States before May 2004 (2000) Notes: The Russian figures are obtained by multiplying the monthly RLMS figures by 12; EU15: Member States belonging to the EU before 1 May 2004. Sources: Russian Federation data are from RLMS rounds 9–12. EU15 data refer to the year 2000 and are from the European Survey of Working and Living Conditions, 2000.

Table 4.1 Costs of absenteeism due to illness in the Russian Federation

Total wage loss Total wage loss Total production (GDP) Total production (US$ billion) as % of GDP loss (US$ billion) loss as % of GDP

2000 40.33 0.55% 97.38 1.34% 2001 52.01 0.68% 105.17 1.37% 2002 56.62 0.71% 104.03 1.30% 2003 60.96 0.71% 112.87 1.31%

Note: The annual average missed days in the Russian Federation are obtained by multiplying the RLMS monthly figures by 12. For details of the calculations see the Appendix. Source: Calculations based on RLMS absenteeism data.

Absenteeism as such is a somewhat crude indicator of the effect of ill-health on the labour market, as is this valuation method, which neglects many other ways in which ill-health has an impact on the labour market. Nor does this method claim to demonstrate a causal relationship. The following subsections use more structural analyses.

4.1.1.2 The impact of health on labour supply and labour productivity

This subsection examines the impact of ill-health on the labour supply and on labour productivity among jobholders in the Russian Federation. (There are a number of methodological challenges involved in properly analysing the issue. Box 4.1 and the Appendix provide technical details.) We also explore the role of health in determining participation in the labour market.

Significant research3 has explored the labour market impact of ill-health in high- income countries. This research demonstrates a negative impact of ill-health both on labour productivity and on labour supply. Mitchell and Burkhauser (1990) used the United States Survey of Disability and Work in 1978 to find that arthritis reduced wages by 27.7% for men and 42.0% for women. Moreover, it reduced the number of hours worked by 42.1% and 36.7% respectively, for men and women. Stern (1996), using the United States Panel Study on Income Dynamics of 1981, showed that limited ability to work due to illness reduced wages by 11.7% and 23.8% for men and women, respectively, when a selection correction for participation in the labour force was introduced. In addition, the probability of staying outside the labour force increased by an estimated 13%. Using the same data, Haveman et al. (1994) estimated that (lagged) ill-health decreased worked hours by 7.4%. Berkovec and Stern (1991), using data from the National Longitudinal Survey of Older Men (1966–1983), found that poor health status reduced wages by 16.7%. Baldwin, Zeager and Flacco (1994), using data from the Survey of Income and Program Participation of 1984, found that health limits reduced wages by 6.1% for men and 5.4% for women. While the varying percentages from these studies lead to theoretical ambiguity, at least in high-income countries there is overall more evidence of a significant negative impact of ill-health on labour supply than on productivity (i.e. wage rates).

Among jobholders, ill-health appears to have had a significant and sizeable impact on labour productivity – but less so on labour supply – in the Russian Federation in recent years. The impact also seems to be more pronounced among males than females. These findings, while slightly different from some in Organisation for Economic Co-operation and Development (OECD) countries, are not necessarily surprising, since the social welfare system in the

Economic impact of health in the Russian Federation 15

3 For an extensive review see, for example, Currie and Madrain (1999) and Suhrcke et al. (2005).

Russian Federation operates very differently than those in OECD countries, affecting the relationship between health and the labour market. In fact, the finding of a significant impact on the wage rate rather than on hours worked is evidence of health’s particularly strong economic impact. (Subsection 4.1.1.3 presents evidence of the existence of one labour supply effect of, in particular, chronic illness and its impact on early retirement.)

The fact that the results obtained using the different methods are qualitatively similar tends to confirm the validity of our findings. We used various methods to develop a sufficiently robust and reliable picture of the labour market impact of adult health. Each method has its own way of addressing the methodological challenges involved in the analysis. In choosing the different approaches we have been guided by relevant literature. Details of our methodology and results are in Box 4.1.

Economic impact of health in the Russian Federation16

Choosing methodologies is largely determined by data availability and by the

informed evaluation of the importance of the endogeneity problem, which tends to

afflict many, if not all, efforts to establish a causal relationship in economic and social

empirical research. In the context of this study the endogeneity problem means that

there could be a simultaneous relationship between the chosen health proxy and

labour market outcomes that would bias the statistical relationship that would be

measured using the most standard econometric technique (i.e. ordinary least squares

estimation). The proposed solutions to the endogeneity problem also critically depend

on the health indicator used and the potential measurement error associated with the

given health indicator, because in some cases the particular kind of measurement

error can offset the bias resulting from the endogeneity problem.

We have used three methods, all adopted from the existing literature. The main data

source to which we applied the methods is the RLMS, specifically the four years from

1999 to 2002. We have also applied the second method (instrumental variable

estimation) to the NOBUS household survey, which has been carried out only once:

in 2003. As health proxy, we used a self-rated health indicator, medically diagnosed

diseases, or work days missed owing to illness.

1. Ordinary least squares (OLS) regressions

This approach is based on a seminal paper by Bartel and Taubman (1979) that uses

a Mincerian wage equation by adding to the usual variables (age, work experience,

Box 4.1 Technical details and results of the impact of ill-health on labour supply and productivity

➤➤

years of schooling, family background) some indicators of diseases, both physical

and mental (heart disease and hypertension, psychoses and neuroses, arthritis,

bronchitis, ulcers, diseases of the nerves, diseases of the liver and bone diseases).

In particular, Bartel and Taubman analyse the effects of such diseases on the basis

of their year of onset in order to disentangle short-term from long-term effects.

We performed a similar exercise by regressing wage rates (in natural logarithms and

at 2000 prices) and the number of hours worked per week (in natural logarithms) on

a large set of the individual-specific health and non-health variables and environmental

variables (Table A.2 in the Appendix lists these variables). The assumption in this

approach – corroborated by a number of statistical tests – is that endogeneity does

not really matter given the specific health indicators used, justifying the use of OLS.

Table A.3 and Table A.4 in the Appendix report the results of four models,

respectively, that differ by date of medical diagnosis for diabetes, heart attack, stroke,

TB and hepatitis (our dataset has data for only these diseases). We find that, as

expected, lung, kidney and spine chronic diseases reduce the wage rate (and hence

productivity). Surprisingly, chronic lung disease increases labour supply. Recently

diagnosed heart attacks and TB reduce wage rate, as expected. Hepatitis diagnosed

very early reduces labour supply while recently diagnosed TB increases it. Indeed,

respiratory and lung-related diseases (such as asthma and bronchitis) seem to have

a positive effect on labour supply. Given the fact that respiratory diseases cause

relatively little limitation on work, a possible hypothesis explaining this puzzle could be

that individuals seek to increase their revenue to compensate for their additional

medical care costs.

Although this approach has been used in the literature, its underlying assumptions are

controversial. The next two methods address endogeneity more directly.

2. Instrumental variables (IV) estimation

When endogeneity is explicitly taken into account, a “simultaneous equation” or

“instrumental variables” approach is typically the preferred option. Following this

method, the endogenous variable (here, the health indicator) should be substituted

by the predicted values coming from its own regression over a set of instrumental

variables plus all the exogenous variables that are part of the model. The researcher

must choose as instruments one or more variables that are correlated with the

endogenous variable but uncorrelated with the error term. The predicted values will

then contain part of the information from the original variable, but they will be purified

from the correlation with errors. This approach was applied to both the RLMS and

the NOBUS data. Since the surveys differ, the precise specification of the estimation

methodology also differs slightly.

Economic impact of health in the Russian Federation 17

➤➤

RLMS

We used individual self-reported health status as the health proxy in the first set of

regressions and the reported number of work days missed due to illness in the

second. The latter is also self-reported, so it may be affected by measurement errors

that are also systematically related to the individuals’ characteristics. We used this

indicator because it could be a more specific indicator of work limitations than the

overall health status. Schultz and Tansel (1995) used the same indicator in another

national context, interpreting it as an “objective” measure of health status. We have

performed two kinds of estimations and both follow Stern (1989) in the choice of

instruments. Stern used medically diagnosed diseases to instrument for self-reported

health indicators.

The variables in the third column of Table A.2 in the Appendix are used as

instruments for respectively self-evaluated health status and missed days due to ill-

health (the chosen date of diagnosis for the last five is between 5 and 10 years

before the interview).

Table A.5 and Table A.6 report estimates for both the logarithm of wage rate and

labour supply, separately by gender. Both health indicators negatively affect the wage

rate, but they do not have a significant influence on labour supply. Reported good health

status increases the wage rate by 22% for women and by 18% for men, compared

to those who were not in good health. Similarly, a work day missed due to illness

reduces the wage rate by 3.7% in the male subsample and by 5.5% among females.

The Sargan test of overidentification (Sargan 1958) does not reject the hypothesis of

exogeneity of the selected instruments. Although this result must be interpreted only

as an indication of exogeneity, because the Sargan test has only little power, it does

support the Bartel and Taubman (1979) assumption of exogeneity of the health

conditions they used in their OLS analysis.

NOBUS

We used the NOBUS4 survey exclusively for the instrumental variable procedure, and

we again used the self-reported health status as a health proxy: the dummy

healthGOOD comprises both “excellent” and “good” self-rated health status (as in the

RLMS analysis). We used a two-stage least squares (2SLS) regression of the logarithm

Economic impact of health in the Russian Federation18

Box 4.1 (cont.)

4 While the RLMS has certain advantages – in particular that it is repeated annually, allowing comparisons over time – the NOBUS survey, so far only held once (in 2003), covers a far larger share of the population – about 44 500 households – and is representative both nationally and for 46 of the larger subjects of the Russian Federation. It captures a range of aspects of household welfare and has a strong focus on household access to social services. Its health component is, however, very small compared to that in the RLMS and hence a direct comparison with the results from the RLMS reported above is not possible.

➤➤

of monthly wage rate and the logarithm of worked hours per week respectively on

age, gender, number of children, private sector employment, secondary school and

university, length of work experience, location indicators and urban/rural indicator.

Secondary school and university are represented by the values 2 and 3 of the

categorical variable schooling derived from a NOBUS categorical variable that is

ordered in 8 levels. Work experience length comes directly from a NOBUS categorical

variable ordered in 5 levels. The urban indicator assumes value 1 for all places with

more than 20 000 inhabitants. We included one location indicator for each region.

Individual health status was instrumented by the parents’ health status. This may be

justified because many chronic diseases are transmitted intergenerationally – either

biologically or socially. Therefore, parents’ health can be correlated with the health of

their offspring without necessarily being correlated with the children’s individual-

specific omitted variables absorbed by the error term. This choice, determined by

data availability, meant that we had to limit our analysis to the subsample of

jobholders who lived in households with their parents. Clearly, this might cause a

selection bias, which is not easy to address.

The results in Table A.7 and Table A.8 in the Appendix show that health has an

impact on wages more than labour supply (among individuals who participate into

the labour force). In particular, males in good health earn about 30% more than

others (i.e. males with fair, bad and very bad health), and females earn 18% more.

The Sargan tests reported at the bottom of Table A.7 and Table A.8 generally support

our choice of instruments (especially for females). We have tried other instruments,

such as location indicators or the number of inhabitants to capture differences in the

prevalence of communicable diseases, differences in the availability of medical

facilities, differences in the prices of health inputs, and differences in environmental

conditions. All these instruments were rejected by the Sargan test. Also, including

parents’ age in addition to parents’ health status increased the probability of

instruments’ endogeneity.

Despite the positive signal previously offered by the Sargan test, concerns remain

about the actual exogeneity of the chosen instruments. For instance, it seems

reasonable to think that high levels of labour supply may increase the probability of

stomach diseases and hypertension, because of the prolonged stress. Moreover, one

may think that heart attacks, strokes or chronic heart diseases are linked to possible

individual risky lifestyles (smoking, drinking, little physical exercise), which may be

correlated with individual-specific error components. To address these concerns, we

have moved from cross-sectional to panel analysis in the next approach.

Economic impact of health in the Russian Federation 19

➤➤

3. Panel regressions

In this third approach we exploited the longitudinal dimension of the dataset by using

panel regression methods. Few studies on the relationship between health and

labour market outcomes have explicitly adopted panel data estimators. Recently,

Pelkowski and Berger (2004) studied the impact of health on employment, wages

and hours worked, distinguishing between temporary and permanent impairments by

using fixed effects estimators. Here, we have followed another recent study, which

makes extensive use of panel data analysis (Cotoyannis and Rice 2001). The authors

suggest the use of Hausman-Taylor (HT) estimators (Hausman and Taylor 1981). In

terms of our previous problem of finding “good” instruments, the main advantage of

this procedure is that it does not require finding valid instruments outside the model,

because it uses the already-included exogenous variables to instrument the relevant

endogenous variable. The only requirement is the inclusion of both time-varying and

time-invariant variables, each of which has to be separated into exogenous and

endogenous ones. Moreover, HT estimators have the advantage over the usual within

(fixed effects) estimators of allowing the effects of time-invariant variables to be

consistently estimated. The disadvantage lies in the strong exogeneity assumptions

to ensure consistency. For this reason, as in Cotoyannis and Rice (2001), we test

such exogeneity assumptions by means of a (Hausman 1978) test. Moreover, to

further improve the precision of our estimates, we also apply the Amemiya and

Macurdy (AM) (1986) estimators, which share the same spirit as HT, but make use of

a more efficient set of instruments (essentially transformations of the HT instruments).

A Hausman test between HT and AM estimators favoured the latter.

To perform our study, we employed the sample of all individuals who have been

followed in rounds 9–12 and who provided answers to all the questions in the survey

we used. This means we can only consider the subsample of jobholders. Owing to

attrition and the relatively high frequency of missing responses, the subsample of

males has only 274 individuals, each observed four times, while the subsample of

females has 476 individuals. To address the problem of an eventual selection bias,

we performed similar estimations, whenever possible, on a significantly larger

unbalanced panel, which, to our comfort, produced similar results.

The results are in Tables A.9–A.12 in the Appendix. In general, we found that good

health status increases wage rate for males, while it does not substantially affect labour

supply. This result is in line with what is obtained in the cross-sectional instrumental

variables estimators of the preceding subsection. However, now the effect of good

health is reduced: being in good health increases the wage rate by about 7.5%.

Surprisingly, good health does not have an impact either on wage rate or labour

Economic impact of health in the Russian Federation20

Box 4.1 (cont.)

➤➤

4.1.1.3 The impact of chronic illness on early retirement

This subsection looks at a very specific potential labour supply effect of health: that of chronic illness on the decision to exit the labour force, to retire. It complements the preceding analysis that also partly looked at labour supply.

Many studies in industrialized countries have shown that ill-health, and in particular chronic illness, affects the decision to exit the labour force: healthy people, other things being equal, tend to retire later than less-healthy ones. Based on a review of various United States studies, Sammartino (1987) concluded that those in poor health are likely to retire between one and three years earlier than those in good health with similar economic and demographic characteristics. Bound, Stinebrickner and Waidmann (2003), based on the analysis of data from the American Health and Retirement Study, estimated that a representative individual in poor health is 10 times more likely than a similar person in average health to retire before becoming eligible for pension benefits. Coile (2003) found that health shocks have a large effect on labour supply decisions by both men and women, mainly when accompanied by major changes in functional status. For example, the onset of a heart attack or stroke accompanied by an important deterioration in the ability to perform “activities of daily living” (e.g. dressing) was estimated to reduce the number of work hours supplied by men per year by 1030 or to raise the probability of leaving the labour force by 42%. A comparable effect of a 654-hour decrease or a 31% increase in the probability of leaving the labour force was found for women.

Turning to evidence from European countries, Jiménez-Martin, Labeaga and Martínez (1999) found that health,5 particularly among men, was a very relevant factor in the decision to retire and for their spouse to retire with them. The authors use information on labour market transitions between 1994 and 1995 from the European Community Household Panel, pooling data from

Economic impact of health in the Russian Federation 21

supply among female workers, unlike what is seen in the cross-sectional instrumental

variables estimations, where the effect on females was even greater than the effect

on male wage rate.

For the sake of completeness, we used an alternative measure of health status: the

“missed days due to ill-health” variable. However, its coefficient was statistically

insignificant both in the wage rate and in the labour supply model.

5 The health variables generally refer to the year 1994 (to minimize the endogeneity bias) and include the following indicators: self-reporting good health, self-reporting a chronic physical or mental health problem (data available only for 1995), having been admitted as an inpatient during the previous year, having visited a doctor between one and five times in the year, and having visited a doctor more than five times in the year.

across the EU, to analyse retirement patterns of individuals and couples in a sample of men older than 54 years and women older than 49. Strong evidence of the influence of health status in the retirement decision is also found by Siddiqui (1997), using data from the German Socioeconomic Panel looking at men in western Germany who had reached the minimum retirement age (which, given the related policies in the country, is considered to be 58 years).6

Indeed, the degree of disability seems to be the dominant factor explaining early retirement, with the probability of leaving the labour force at the earliest possible age for disabled men being four times that of men without disability. As Siddiqui notes, these results suggest that improving employees’ health could be a highly effective measure to raise the actual age of retirement.

Applying the various approaches used in other countries to the Russian Federation’s case reveals a statistically very robust and sizeable impact of chronic illness on both age of retirement and on the probability of retiring in the subsequent year. We followed two different, complementary approaches: a Cox regression and a panel logit regression. Controlling for other relevant determinants of the decision to retire (e.g. age, gender, income), both approaches confirm the finding that chronic illness increases the probability of retiring. The former approach (Cox regression) assesses the effect of chronic illness on the probability that an individual will retire in a given year after the first year of employment. This methodology’s limitation is that we cannot be entirely sure about the direction of the causality – does ill-health predict retirement or vice versa? The second approach (the panel logit regression) is more appropriate to address this issue, since it examines the effect of chronic illness on the probability of retiring in the subsequent year.

The Cox regression indicates that a hypothetical male aged 55 on median income and having certain other characteristics7 would be expected to retire at age 59, while a chronic illness would lower his expected retirement age by two years (Figure 4.2). While the technical details of the regression results (see Box 4.2) can be difficult to interpret, they are more intuitively understandable if applied to a hypothetical individual. Similar results are obtained for females. Strictly speaking, though, we can only talk about evidence of an association between chronic illness and earlier retirement, since the available data offer no way of discovering the time of onset of a chronic disease for an individual. In particular, we do not know whether the illness occurred before or after retirement, so we cannot say, from this analysis, whether the statistical association reflects the effect of chronic illness on retirement or vice versa. We can, however, address this issue by using a panel logit regression.

Economic impact of health in the Russian Federation22

6 The self-employed are withdrawn from the sample due to their different pension systems. 7 The other characteristics are: married, has one child, not smoking, not drinking, of normal weight, with a high-school diploma, born in the Russian Federation and living in an urban area.

A Cox regression allows us to estimate the precise moment that an event takes place

as time proceeds. It is usually employed in survival analysis, where the outcome

considered is death. It can also serve the purpose of estimating the timing of

retirement. We estimated a Cox regression model on the age at retirement, using

data from the 11th round of the RLMS (2002), where we can find retrospective

information on job retirement.

Estimating a Cox regression model on the age at retirement: this is a model of a

hazard regression where the log hazard function of retirement log[h(t)] is assumed to

be a linear function of a baseline hazard function and the effect of p covariates. Formally:

log[h(t)] = log[h0(t)] + β1x1 + β2x2 + ... + βpxp

Thus, the parameters we estimated represent a proportional shift of the baseline

hazard function due to the covariates. A positive parameter means an increase in the

risk of retiring from work during the overall time period (since first employment). The

results are in Table 4.2. The reported coefficients should be interpreted as follows: a

Economic impact of health in the Russian Federation 23

1.0

0.8

0.6

0.4

0.2

0.0

20 40 60 800

Age

No chronic illness Chronically ill

S ur

vi va

l f un

ct io

Males

Figure 4.2 Probability of remaining in the workforce with and without chronic illness, by age, based on Cox regression model Note: Cox regression results are described in Box 4.2 Source: Calculations based on RLMS round 11.

Box 4.2 Cox regression: technical details and results

➤➤

positive coefficient means an increase in the risk of experiencing the event (retirement from work in this case) and a negative coefficient is associated with a decrease in the risk of experiencing the event. (The test based on Schoenfeld residuals showed that the null hypothesis – the effect of chronic illness on the decision to retire being proportional – is not rejected.)

We controlled for a set of demographic and socioeconomic indicators: age, gender, income, education, etc. The health variable of particular interest is the presence of a chronic illness. A positive coefficient on the chronic illness variable indicates an increase in the probability (i.e. the hazard) of entering retirement, relative to the baseline first year of employment.

Those who are married, widowed or divorced are more likely to retire later from the job market than those who never married. The effect of age is U-shaped. Females retire later but the effect is weak and decreases with age. Smoking increases the risk of retiring, but the effect decreases with age. The effect of weight is interesting: those who are below the normal weight (in terms of body mass index) retire earlier, whereas those who are overweight or obese are more likely to retire later than individuals of normal weight. Reported drinking does not have any significant effect, but chronic illness has a positive and highly significant effect. This means that, after having controlled for other factors, we find that, in contrast with findings from the Kaplan- Meier estimates, those suffering from any chronic disease are more likely to retire earlier from the job market. Moreover, the effect of chronic illness interacts with income: the higher the income level the weaker the effect of chronic illness. In addition, we find that workers below the poverty line retire earlier and that income has a negative effect (i.e. the higher the income level, the later a worker retires). The number of children has no significant effect for males but it has a positive one for females. Finally, the estimates from the Cox model suggest that people born in the Russian Federation are more likely to retire than those born outside the Russian Federation, and those living in a village are more likely to retire earlier than those living in urban areas.

Economic impact of health in the Russian Federation24

Table 4.2 Results of Cox regression model on age to retirement

Variable Coefficient

Age -.492*** Age squared .003*** Female -.423*** Age* female .0132*** Married -.275*** Cohabit -.129 Widowed or divorced -.262*** Chronic illness .228*** Poverty status .495*** Household income -.0116*** Hh income* chronic illness -.014** High-school diploma -.447*** No. of children < 7 years -.123 Female* no. of children under 7 .378*** Born in Russian Federation -.141*** Living in village .113**

Notes: *** 1% significance level; ** 5% significance level; * 10% significance level; “Hh”: Household head.

Box 4.2 (cont.)

The panel logit regression results show that an individual who suffers from chronic illness in one period has a significantly higher probability of retiring in the subsequent year compared to the same individual free of chronic illness. Some of the respondents in the RLMS have been followed throughout several years of the survey.8 This allows us to use a panel logit regression in order to assess the impact of chronic illness in one year on the probability of retirement in the subsequent year. In this case we assess the effects of chronic illness on the probability of entering retirement in the next year, not the effect on the probability of retiring in a given year after first employment. Otherwise, the set of explanatory variables to be controlled for is identical to the Cox model. The results (Table 4.3) show a very similar pattern to those based on the Cox regressions (Table 4.2), with only minor differences. In particular, chronic illness emerges as a highly significant predictor of subsequent retirement. Given the different methodology, this result provides a more reliable basis for claiming causality between chronic illness and the probability of retirement. The magnitude of its effect is large compared to other variables in the model.

In either approach, the effect of chronic illness is found to vary with income: the lower the income the more chronic illness affects the retirement decision.

Economic impact of health in the Russian Federation 25

Table 4.3 Random effects logit regression results

Variable Coefficient

Age -0.492 *** Age squared 0.003 *** Reference: male Female -0.423 *** Age* female 0.013 *** Married -0.275 *** Cohabit -0.129 Widow or divorced -0.262 *** Chronic illness 0.228 *** Poverty status 0.495 *** Household income -0.012 *** Income* chronic illness -0.014 ** High-school diploma -0.447 *** Number of children in household -0.123 Female* no. of children 0.378 *** Born in Russian Federation -0.141 *** Living in village 0.113 ** Constant 4.192 *** Rho 0.141 **

Notes: *** 1% significance level; ** 5% significance level; * 10% significance level.

8 This is the “panel” component of the RLMS, which in principle offers important opportunities for testing hypotheses that involve a causal perspective. One shortcoming of this panel dimension is that it does not feature a true panel design, as households that moved from their dwelling and individuals who moved from their household are not followed. In any case, the effect of attrition is relatively modest and is highest for the respondents from the metropolitan areas of Moscow and St Petersburg.

This implies that less-affluent people carry a double burden of ill-health: first, they are more likely to suffer from chronic illness; second, once they fall ill, they suffer worse economic consequences than rich people – a feature that tends to perpetuate socioeconomic disadvantage.9 Technically speaking, this result is reflected in the statistically significant interaction term between income and chronic illness in the regression models. As far as Cox regression is concerned, this can be illustrated by comparing the effect of chronic illness in the hypothetical individual described above to another individual with the same characteristics but with an income of 50% of the median: he will retire, on average, at 58.8 years without a chronic illness but at 56.3 if a chronic illness is present – 2.5 years earlier, as opposed to 2 years earlier in the case of the richer individual. This result exemplifies the gradient of the impact on the basis of the panel logit model: among males with a very high income, the presence of a chronic illness has no effect on retirement age, while men just below the average of the income distribution have a 24% higher probability of retiring early compared to their healthy counterparts (Figure 4.3).

Economic impact of health in the Russian Federation26

9 Note that we are not able to explore a similar variation of the effect of ill-health across the income scale in the wage and earnings regressions presented above. This would require a different approach, for instance a quantile regression (see, for example, Rivera and Currais (1999) for an application of quantile regressions relating to Brazil).

0.18 0.18 0.21

0.29

0.23

0.40

0.32

0.56

0.43

0.62

>95th income

percentile

75th–95th income

percentile

50th–75th income

percentile

25th–50th income

percentile

<25th income

percentile

RICHEST POOREST

0.7

0.6

0.5

0.4

0.3

0.2

0.1

Not chronically ill Chronically ill

Figure 4.3 Average predicted probability of retiring in the subsequent period, hypothetical male at varying income levels: based on panel logit model Note: Results refer to the hypothetical individual described in the text. Source: Calculations based on RLMS rounds 9–11.

4.1.1.4 The impact of alcohol consumption on the probability of being fired

Heavy alcohol consumption is arguably the most important proximate cause of adult mortality in the Russian Federation. Furthermore, several studies in other developed countries have shown that heavy consumption has a negative impact on earnings, incomes and wages, because it reduces individual productivity and may create problems with working arrangements for the employer.10 In this subsection we apply this idea to the available Russian data by exploring whether alcohol consumption in one year (2001, round 11 of RLMS) increases the risk of job loss in the subsequent year (2002, round 12). The rationale for this exercise is that job loss would be a natural consequence of an appreciable reduction in individual productivity.

We find that one negative economic impact of severe alcohol consumption is that it significantly increases the probability of losing one’s job. Using a panel probit model and controlling for gender, age, education, work experience, wage rate and the ownership type of the employing organization, we find that alcohol has a positive and statistically significant effect on the probability of being fired, even if its size appears relatively small (see Box 4.3). The small size may reflect the simplified structure of the estimated model. Further research would be necessary to disentangle alcohol’s complex but no doubt important effects on the Russian labour market.

Economic impact of health in the Russian Federation 27

10 See e.g. Mullahy (1991) and Cercone (1994).

Box 4.3 Technical details and results of panel probit model on the probability of being fired

We estimated a probit model of the probability of being “fired”, which we made

dependent on gender, age (in months), wage rate, possession of a high-school

diploma, post-secondary years of schooling, work experience, type of enterprise

ownership (state, foreigners, private Russian owners) and, finally, daily alcohol

consumption (in grams of pure alcohol) and squared daily alcohol consumption.

The dummy variable “fired” was defined such that it takes the value 1 if an individual

was employed in round 11 (2002), was not employed in round 12 (2003), and yet

participated in the workforce in round 12. An alternative definition embodying the

condition of being unemployed in round 12 produced a very similar identification.

Through the chosen set-up, we assumed that alcohol consumption had a nonlinear

effect on the probability of being fired. This supposition was confirmed by other

analyses. We applied the Huber/White/sandwich estimator of variance in place of the

traditional calculation to obtain robust standard errors. The results are in Table 4.4.

➤➤

4.1.2 Some wider costs of adult mortality: effect on other household members

So far we have focused on the impact of adult ill-health on the individual directly concerned. This captures only part of the overall effect of adult ill- health, as it leaves out the impact on other people, in particular household members. In this subsection we assess the consequences of an individual’s death on surviving household members. We specifically explore two potentially related types of “consequences” of a household member’s death: depression and alcohol consumption. Both tend to decrease labour productivity and weaken social ties, so they can be interpreted as relevant proxies of economic outcomes.

The death of a household member was found to increase the probability of suffering from depression by 53%. Again, we exploited the panel dimension of the RLMS, i.e. rounds 11 (2002) and 12 (2003), enabling us to assume a more causal interpretation of the results. We included in the sample only those living in households whose composition remained constant between 2002 and 2003 or was altered because one or more members died. This means that we excluded households who lost members for reasons other than death (e.g. migration, new household formation). Using probit analysis and controlling for relevant variables, we explored the effect of a household member who died in 2002 on the probability that any surviving household member would experience depression in the subsequent year (2003). Results are in Table 4.5. As expected, the probability of depression decreases with the age of the deceased. We also controlled for possible differences in per-capita income in order to check whether depression was related to this factor rather than to the death per se. It appears that differences in per-capita income do not affect the probability of depression.

Economic impact of health in the Russian Federation28

Box 4.3 (cont.)

Table 4.4 Panel probit results on alcohol as a determinant of being fired

Variable dF/dx x-bar

Gender (male = 1) -.00208 1.54 Age (in months) .00006 ** 472 Monthly normal wage in 2002 roubles -1.53e-06 ** 3422 Secondary school diploma (if yes = 1) -.0043 1.14 Post-secondary years of schooling -.0011 ** 3.28 Number of working years -.0010 *** 19.03 Publicly owned firm -.00208 0.68 Foreign-owned firm .00852 0.05 Privately owned firm .00508 * 0.43 Alcohol consumption (per week in grams of pure alcohol) .00030 ** 15.6 Alcohol consumption squared -2.84e-06 ** 1818

Notes: *** 1% significance level; ** 5% significance level; * 10% significance level; dF/dx is for discrete change of dummy variable from 0 to 1; z and P > |z| are the test of the underlying coefficient being 0; Number of obs. = 4173; Wald chi2(11) = 60.89; Log likelihood = -311.60966.

Table 4.5 Regression results on the effect of a household member’s death on depression

Variable dF/dx x-bar

Gender (male = 1) -.00208 1.54 Age (in months) .00006 ** 472 Wage rate in 2002 roubles -1.53e-06 ** 3422 Secondary-school diploma (yes = 1) -.0043 1.14 Post-secondary years of schooling -.0011 ** 3.28 Number of working years -.0010 *** 19.03 Publicly owned firm -.00208 0.68 Foreign-owned firm .00852 0.05 Privately owned firm .00508 * 0.43 Pure alcohol consumption in grams per week .00030 ** 15.6 Alcohol consumption squared -2.84e-06 ** 1818

Notes: *** 1% significance level; ** 5% significance level; * 10% significance level; Number of obs. = 8113; LR chi2(9) = 321.50; Log likelihood = -3740.8969.

Alcohol consumption was found to increase by about 10 g per day as a consequence of a death in the household. If the deceased was employed, then the survivor’s alcohol consumption increased 25 g per day. Using the same two years, we employed a tobit model including essentially the same control variables as in the depression model. Surprisingly, if the deceased was the household head, there was no independent impact, at least not in the short term that we examined. Detailed results are in Table 4.6.

Table 4.6 Regression results on alcohol consumption in response to a household member’s death

Variable Coefficient

Gender (male = 1) 36.47 *** Age (in months) -0.01 *** Employed (yes = 1) 23.21 *** Difference in per-capita income (after and before death) 0.0005 *** High-school diploma (yes = 1) 10.75 *** Number of deceased household members throughout the past year 10.55 ** Number of deceased household members who were household heads 4.40 Number of deceased employed household members 25.19 * Constant -44.95 ***

Notes: *** 1% significance level; ** 5% significance level; * 10% significance level; Number of obs. = 8170; 3677 left-censored observations at alcohol<=0; 4493 uncensored observations; LR chi2(8) = 1002.07; Log likelihood = -26843.276.

4.1.3 The effect of chronic illness on income

Chronic illness has had a negative impact on household incomes in the Russian Federation, particularly in the period 1998–2002. In order to deal with some technical constraints on estimating the causal effect of health on economic outcomes – mainly the issue of endogeneity of the health proxy used – we employed a strategy that differs from that used in our other analyses for the

Economic impact of health in the Russian Federation 29

present study.11 We used a difference-in-differences estimator combined with a propensity score-matching technique, applied to the RLMS surveys from 1994 to 2002. Essentially this technique allowed us to compare pairs of households that were identical except for the presence of health problems. Details of the methodology and the results are in Box 4.4.

Using a two-step procedure, we find chronic illness to contribute to an annual loss of 5.6% of per-capita median income for a hypothetical individual with given characteristics.12 The first step confirmed a negative effect of poor health (in general) on household income. This effect is greater in 1998–2002 than before the Russian financial crisis. We then used a more detailed logit model to assess the extent to which chronic illness increases the likelihood of experiencing adverse health events. These steps show that chronic illness increases the risk of health problems. Combining the effect of chronic illness and poor health on income then gives the overall indirect impact of chronic illness on household income.

Economic impact of health in the Russian Federation30

11 In the previous sections we tried to address endogeneity either by exploring the lagged effect of ill-health on a specific economic outcome using panel regressions or by applying an instrumental variable estimation in the cross-section regressions. (In one case we also used the instrumental variable estimation in the panel context.) 12 The household characteristics are: in urban areas, with no smokers and no ex-smokers, no people aged over 60 or below 14, with at least two workers and at least one person who has a high-school diploma.

Box 4.4 Technical details and results of household income impact

To address the endogeneity problems involved in estimating the effect of health on

economic outcomes, we used a strategy that does not employ instrumental variables.

A difference-in-differences estimator combined with a propensity score-matching

technique is described in Rosenbaum and Rubin (1983) and Heckman, Ichimura and

Todd (1997). With this approach, every household experiencing a health problem is

matched to a similar household not having health problems. Similarity is defined in terms

of a propensity score, i.e. the propensity of experiencing an adverse health event given

the household characteristics (for instance whether the household members suffer from

chronic illness). By comparing the experiences of two similar households in this way, we

can identify the causal effect of health on income. The logic is essentially that of comparing

two groups that differ only in relation to the variable of interest. This strategy makes it

possible to separate the impact of individual health from other contingent effects.

The results (Table 4.7) show the effect on total income of two different events related

to poor health: generic health problems and hospitalization. We devised two separate

estimates for the periods 1994–1998 and 1998–2002 to capture the differences

between the period immediately before the economic crisis, which began in 1998,

and the period after the crisis start. The results irrefutably confirm a negative effect of

poor health on household economic well-being: the effect is greater in the later period. ➤➤

We have so far demonstrated various channels through which health has had an impact on various economic outcomes in the Russian Federation. This is in line with findings from an increasing body of literature on health and the economy in other countries, both wealthier and less affluent. In each estimate presented here, the results proved statistically highly significant, and where size could be assessed, it is notable.

Section 4.2 looks ahead and asks, “what would the economic benefits to the economy be if the adult disease burden due to NCDs and injuries were reduced by a certain extent over a defined period of time?”.

Economic impact of health in the Russian Federation 31

To estimate the specific impact of chronic diseases we used a logit model to assess

whether and to what extent chronic illness increases the likelihood of experiencing

adverse health events. The corresponding results are not reported here, but are

available from the authors. The results show that chronic illness increases the risk of

health problems as well as of hospitalization and of undergoing a surgical procedure.

Our results confirm that chronic illness does indirectly and negatively affect the

economic well-being of Russian households, especially since the economic crisis in

1998. But what can we say about the magnitude of the effect? It is not possible to

provide a comprehensive answer since the risk of health problems depends not only

on the presence of chronically ill people in the household, but also on other factors:

number of smokers, household size, number of older people, etc. However, we can

provide a specific answer for a specific population: households in urban areas, with

no smokers or ex-smokers, no one over 60 or below 14, with at least two workers

and at least one person who has a high-school diploma. For this restricted population

the average difference in the probability of having health problems between households

with chronically ill members and those without is 0.219. The difference in the

probability of being hospitalized is 0.038, and the difference in the probability of

undergoing a surgical operation is 0.018. Multiplying these differences by the effect of

health problems, hospitalization and surgical operation on economic outcomes gives

the indirect effect of chronic illness on income. The effect corresponds to 5.6% of

median per-capita income.

Table 4.7 Results from difference-in-differences estimator combined with propensity score technique: effect of adverse health on total income for different periods

Total income

1994–1998 1998–2002 1994–2002

Health problems -22.255 -135.98*** -83.147*** Hospitalization -136.19*** -105.83*** -82.30***

Note: *** 1% significance level.

4.2 If health were improved, what macroeconomic benefits would result?

Here we estimate the macroeconomic benefits of reducing mortality rates due to NCDs and injuries among Russian adults and find that they would be substantial, irrespective of the evaluation method. The substantial and certain economic benefit would occur despite the fact that we focus only on the effect of mortality reductions, setting aside the additional impact of the potential associated morbidity reduction. Our main findings are detailed below.

• The static economic benefit – valuing a life year lost by one GDP per capita – of gradually bringing the Russian Federation’s adult NCD and injury mortality rates down to current EU15 average rates by the year 2025 is estimated to be between 3.6% and 4.8% of the 2002 Russian GDP.

• The broader “welfare” benefits – valuing a life year by a more broadly defined “value of life” estimate – of achieving current EU15 average rates by 2025 are estimated to be as high as 29% of the 2002 Russian GDP.

• The dynamic benefits, i.e. the effect on economic growth rates, are massive and growing over time. Even if the future returns are discounted to the starting-year value (2002), they represent a multiple of the static GDP effects.

This section proposes different ways of looking at the country-wide impact of health on the Russian economy. We distinguish between static (Section 4.2.1) and dynamic cost estimates (Section 4.2.2). The static cost estimates serve an illustrative purpose and are easier to calculate, while the slightly more complex dynamic cost estimates that assess the impact of health on economic growth present a more complete macroeconomic impact assessment and should be of greatest interest to policy-makers. We also look at the static welfare benefits to do justice to the fact that quality of life is more important than the quantity of goods produced.

4.2.1 The benefits of reducing NCD and injury mortality: a simple static calculation

The first step in evaluating the economic benefits of future mortality scenarios is to define the mortality scenarios themselves. To do so, we followed a deliberately simple approach. We defined three different scenarios for the future development of adult mortality (ages 15–64) between 2002 (the most recent year for which the WHO Mortality Database provides data) and 2025. Focusing only on the impact of NCDs and injuries, we included only changes in adult mortality that are driven by the evolution of adult NCD and injury

Economic impact of health in the Russian Federation32

mortality rates.13 Hence, our starting point is the definition of three NCD and injury mortality scenarios. Once we define the starting conditions for the models, we can assess future trends by inserting estimated changes in adult mortality rates in the different scenarios.

Scenario 1: Optimistic scenario

This scenario assumes that policies are adopted that bring about a decline in Russian mortality rates from NCDs and injuries to the most recent available level for the 15 countries belonging to the EU prior to May 2004. This corresponds to an annual rate of reduction of 4.6% for NCDs and 6.6% for injuries.

Scenario 2: Intermediate scenario

This scenario assumes that policies are adopted that achieve half the improvement seen in the optimistic scenario. It assumes an annual reduction of 2.3% for NCDs and 3.3% for injuries.

Scenario 3: Status quo

In this scenario, the 2002 level of adult mortality rate from NCDs and injuries in the Russian Federation is assumed to remain constant until 2025. One might object that this is unnecessarily pessimistic as NCD and injury mortality are expected to decline simply as an almost automatic response to the very positive recent (and perhaps future) economic development record, even if no specific additional efforts are undertaken to improve adult health. The future trends cannot be known, but Figure 4.4 shows that (a) the long- term increasing trend of noncommunicable (in particular cardiovascular) disease mortality and injury-related mortality over the past decades leaves very limited hope for a sudden or even gradual reversal of these trends, assuming no change in health/economic policy; and (b) these cause-specific mortality rates have increased significantly in recent years, despite particularly strong economic growth. For these reasons, a scenario where the relevant cause- specific mortality rates would remain at their 2002 levels is considered modestly optimistic.

The effect that each of these scenarios would have on adult mortality rates is illustrated in Figure 4.5, assuming other cause-specific adult mortality rates remain constant.

Economic impact of health in the Russian Federation 33

13 In doing so, we are understating the broader health impact that eventual broad-based health interventions are likely to have.

Economic impact of health in the Russian Federation34

100

200

300

400

500

600

700

800

900

1000

20 02

20 03

20 04

20 05

20 06

20 07

20 08

20 09

20 10

20 11

20 12

20 13

20 14

20 15

20 16

20 17

20 18

20 19

20 20

20 21

20 22

20 23

20 24

20 25

Scenario 3

Scenario 2

Scenario 1

Benchmark EU15 (2001)

Figure 4.5 Three scenarios for Russian adult NCD and injury mortality rates (2002–2025) and those of the EU Member States before May 2004 (2001) (age 15–64, per 1000) Notes: EU15: Member States belonging to the EU before 1 May 2004; Scenarios are based on the assumptions outlined in the text. Source: Table 4.8 for initial values and EU15 benchmark.

350

19 85

19 86

19 87

19 88

19 89

19 90

19 91

19 92

19 93

19 94

19 95

19 96

19 97

19 98

19 99

20 00

20 01

20 02

300

250

200

150

100

CVD

External causes

20 03

Figure 4.4 Standardized death rates due to CVD and external causes in the Russian Federation (age 0–64, per 100 000) Note: CVD: cardiovascular disease. Source: WHO Regional Office for Europe European Health for All database 2006.

None of these scenarios is based on the detailed modelling of the impact of specific policy interventions; this remains a topic for further research. All that matters for our purposes in this study is that the chosen scenarios can be considered to be plausible, i.e. based on the mortality reductions that other western or northern European countries have achieved over the decades. While the most ambitious scenario is indeed very ambitious, it has been achieved in the past (for example the North Karelian and Finnish experiences (World Bank 2005)). Readers will note that the actual levels of the mortality rates foreseen in each of these scenarios are less relevant than the difference between them: the difference between any two scenarios determines the opportunity cost or benefit of the respective scenarios.

Table 4.8 shows the actual mortality rates from NCDs and injury in the Russian Federation (2002) and in the 15 countries belonging to the EU prior to May 2004 (2001). We added CVD mortality for illustrative purposes, as it accounts for the greatest share of total adult NCD mortality. Russian rates are a multiple of the European ones, and the difference is particularly great in the case of CVD mortality.

Next, we undertook a basic economic evaluation to explore the effects of the different scenarios in relation to potential policies to reduce NCDs and injuries up to the year 2025. To evaluate the different scenarios, we use first a “narrow” approach – using foregone production, i.e. GDP per capita – and then a broader perspective to capture the value of living longer without illness.

In what follows we illustrate different ways of assessing the static economic benefits of pursuing the most optimistic and intermediate scenarios compared to the status quo scenario. The first method uses the value of per-capita production (GDP per capita) as the value of a year of adult life lost. This is admittedly a crude (as it lacks a profound theoretical basis), yet not uncommon,14 way of valuing mortality reductions. The second approach rests on a sound theoretical welfare basis, recognizing that the true cost of a year of life lost greatly exceeds foregone output.

Economic impact of health in the Russian Federation 35

Table 4.8 Cause-specific adult death rates in the Russian Federation and EU Member States before May 2004 (age 15–64, per 100 000)

Russian Federation EU15 Death rates as a % of EU15

Noncommunicable diseases 605 206 294% Injuries 281 58 484% Cardiovascular diseases 348 37 941%

Notes: Russian rates refer to 2002; EU15 (Member States belonging to the EU before 1 May 2004) rates refer to 2001 or latest available year; The EU15 average is population weighted. Source: WHO Regional Office for Europe 2006.

14 The Commission on Macroeconomics and Health also applies a version of this methodology (see CMH 2001, p. 103).

Static GDP effects

The static economic benefit of gradually bringing the adult NCD and injury mortality rates down to match the current rates for the 15 Member States belonging to the EU prior to 1 May 2004 by the year 2025 (i.e. the optimistic scenario) is estimated to be between 3.6% and 4.8% of the 2002 Russian

Key facts

Organisation World Health Organization (WHO)

Document type Publications

Adoption date January 1, 2007

Source World Health Organization