InGRID Presentation Gabriele Mazzolini (UNIMIB & UCSC) 22 - - PowerPoint PPT Presentation

SLIDE 1

InGRID Presentation

22 January 2016

Gabriele Mazzolini

(UNIMIB & UCSC)

Creli Centro di ricerche per i problemi del lavoro e dell’impresa

SLIDE 2

Presentation outline

• CRELI – Research Centre in Labour Economics and

Human Resource Management

• MCA – Multiple correspondence Analysis
• Is there a bad effect of work organisation

systems on health?

with E.Cottini and C.Lucifora

SLIDE 3

Who we are

• CRELI - Research Centre in Labour Economics and

Human Resource Management gathers researchers in the broad fields of labour economics, human resource management and health and safety at work, with expertise both in theory and applied research.

• We have gained reputation for econometric analysis
• f large micro data

– Census data – Panel data surveys – Administrative archives – time-use diary data

SLIDE 4

Who we are

• The main research topics of interest are:

– Work organization in firms – Personnel economics: pay, incentives and careers within firm – Entrepreneurship – Gender pay differentials, occupational segregation and equal

• pportunity policies

– Job quality – Health and safety at work – Schooling, training and investment in human capital – Wage determination, wage differentials and mobility – Labour market regulation – Industrial relations – Employment policies and welfare – Migration and labour market effects – Public and Private sector labour markets – Regional labour market and regional disparities

SLIDE 5

Who we are

• The Centre has been leading research at the

European level on topics related to work and labour market, within the EU Framework Programme such as:

• LoWER, European Low-wage Employment Research

network;

• EQUALSOC, Economic Change, Quality of Life and Social

Cohesion;

• Health at Work, Health Economics and Working

conditions;

• LEED , Linked Employer Employee Data;
• GINI – Growing Inequalities;
• ERFI, Institut national d’etudes demographiques.

SLIDE 6

My research interest

• Occupational health and safety
• Flexicurity and the role of institutions in the

labour market

• The entry of young people in the labour

market

• The impact of tax audit on labour

SLIDE 7

Multiple Correspondence Analysis

• Multiple correspondence analysis (MCA) is a data

analysis technique for nominal categorical data, used to detect and represent underlying structures in a data set.

• It represents data (outcomes of variables) as points

in a Euclidean space.

• MCA can be viewed as a generalization of PCA where

the variables are categorical, not continuous.

SLIDE 8

Multiple Correspondence Analysis

• MCA performs a PCA on an indicator or a Burt matrix;

it explores the relationships within a set of variables.

• Associations between variables are uncovered by

calculating the inertia (the weighted variance of the coefficients) of the indicator or the Burt matrix

• These associations are then represented graphically

as "maps", which eases the interpretation of the structures in the data.

SLIDE 9

MCA: An application (1)

• Suppose we run a MCA using some questions from the 2008

European Value Study survey to identify different group of voters in Italy and in Belgium

• Q1: Would you like to have immigrants as neighbors? Yes/No

– 2 possible outcomes - Dummy variable

• Q2: Why are there people in this country who live in need? Choose

between the following four possible reasons

• because they are unlucky
• because of laziness and lack of willpower
• because of injustice in our society
• it’s an inevitable part of modern progress

– 4 possible outcomes - Categorically distributed variable

• Q3: Please tell me whether you think it can always, sometimes or

never be justified paying cash for services to avoid taxes

• 5 possible outcomes – Discrete variable

SLIDE 10

MCA: An application (2)

• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

SLIDE 11

MCA: An application (2_1)

immigration_yes immigration_no

• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

SLIDE 12

MCA: An application (2_2)

poor=progress poor=unlucky poor=system poor=lazy

• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

SLIDE 13

MCA: An application (2_3)

never_black preferebly_no_black black_happens almost_ok_black

• k_black
• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

SLIDE 14

MCA: An application (3)

immigration_yes immigration_no poor=progress poor=unlucky poor=system poor=lazy never_black preferebly_no_black black_happens almost_ok_black

• k_black
• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

SLIDE 15

MCA: An application (4)

immigration_yes immigration_no poor=progress poor=unlucky poor=system poor=lazy never_black preferebly_no_black black_happens almost_ok_black

• k_black
• 3
• 2,5
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 2
• 1

1 2

Europeanism Anti Europeanism Conservative Progressive

PD

Forza Italia NCD Lega Nord M5S SEL

SLIDE 16

immigration_yes immigration_no poor=progress poor=unlucky poor=system poor=lazy never_black preferebly_no_black black_happens almost_ok_black

• k_black
• 2
• 1,5
• 1
• 0,5

0,5 1 1,5 2

• 3
• 2
• 1

1 2 3

MCA: An application (5)

Authoritarian Liberal Conservative Progressive VB SP.A CD&V ULD Groen LDD

SLIDE 17

Is there a bad effect of work

• rganisation systems on health?

Elena Cottini

(UCSC)

Claudio Lucifora

(UCSC and IZA)

Gabriele Mazzolini

(UNIMIB & UCSC)

Creli Centro di ricerche per i problemi del lavoro e dell’impresa

SLIDE 18

• “Workmen […] when they are liberally paid

by the piece, are very apt to overwork themselves, and to ruin their health and constitution in a few years” (A. Smith, 1776, p.83).

SLIDE 19

Introduction

• Substantial changes in the organisation of labour

processes associated to:

• product market competition
• technological progress
• erosion of union power
• innovative management practices
• What we observed in the last decades was that

these changes contributed to increase firm performance and (also) workload demands with a larger use of HIM practices (Bloom and van Reenen 2010, for a review)

SLIDE 20

Introduction

• Evidence for EU countries suggests employees are

exposed to high work intensity (Eurofound, 2005):

– high speed (25 per cent reporting to be exposed to almost constant high working speeds) – tight deadlines (30 percent) – shortage of time at work (22 percent) – long hours (20 percent)

• What are the implication of the increase in

work intensity on worker’s health?

• Available evidence suggests that riskier and more

stressful jobs are associated with poorer health (Oecd, 2008; Bardasi and Francesconi, 2004; Green, 2006; Caroli and Bassanini, 2014)

SLIDE 21

This paper

• we present cross-country evidence, on the links

between health at work and work organisation systems in Europe using the 2000 and 2005 waves of the European Working Conditions Survey (EWCS).

• we complement existing evidence as follows:

₋ health at work indicators

₋ single indicators of mental and physical health

₋ management practices

₋ no ex-ante classification of single work practices, we use multiple correspondence analysis to see how practices bundle together (i.e. Traditional, Taylorism, Lean Production and Discretionary learning)

₋ distributional effects of different management practices

SLIDE 22

Review

Work organisation and health

• the adoption of work practices can have mixed

effects on physical and mental health:

(+) autonomy, teamwork, rotating tasks and multi- skilling processes, could enrich the working environment and improve well-being (-) intensified work pace, strict supervision, increased demands and work intensity could increase work strain and stress

SLIDE 23

Review

• Several papers in the occupational health literature

focused on the association between single practices and specific health risks:

– physically demanding tasks, ergonomics and musculoskeletal disorders (Buckle 2005) – work intensity, stress and risk of cardiovascular disease (Chandola, 2008) – monotonous, repetitive tasks and mental health, anxiety

• r depression (Cottini and Lucifora, 2013)

– long working hours, night shifts and and sleep-related problems (Harrington 2001)

• Also papers in the health economics literature:

– Booth and Francesconi 2002, Llena-Nozal 2009; Robone et al. 2011; Cottini and Lucifora 2014; Cottini 2009; Serrano and Cabral 2005 ; Caroli and Bassanini, 2014

SLIDE 24

Work Organisation systems

• Work practices never come on their own but in bundles of

coherent work organisation systems. (Macduffie 1995, Macduffie and Pil 1997).

– i.e. work formalisation and constraints on pace of work are linked to repetitive tasks, while job autonomy is usually linked to learning new things as a consequence of autonomous problem-solving

• Among the most referred work organization systems are:
• Traditional
• Taylorism
• Lean production
• Discretionary learning
• Only few studies have investigated the effects of work
• rganisation systems on different employees’ outcomes,

and the majority of these studies refer to case studies, single industries or plants (see Hasle et al, 2012 for a review on the manufacturing industry).

SLIDE 25

Data

• European Working Conditions Survey (EWCS)

provides information on:

• Physical and mental health problems at the workplace
• Work organization/management practices
• Pay for performance
• Standard demographics and firm characteristics
• Our sample includes workers (aged 15-64) in

private firms from EU-15 in 2000 and 2005 waves.

SLIDE 26

Measuring Health

• Does your work affect your health, or not? If yes, how

does it affect your health?

• 1. Skin problems; 2. Respiratory problems;
• 3. Stomach-ache; 4. Hearth disease; 5. Stress;
• 6. Sleeping problems; 7. Anxiety; 8. Irritability
• Out of the above indicators we construct two

synthetic variables:

• Physical health dummy: takes value 1 if at

least one of the health problems (1 to 4) above has been mentioned, 0 otherwise (physich_dum).

• Mental health dummy: takes value 1 if at least
• ne of the health problems (5 to 8) above has

been mentioned, 0 otherwise (mentalh_dum).

SLIDE 27

MCA for work organization

• We follow Arundel, Lorenz, Lundvall and Valeyre

(2006). Their methodology consists of two steps:

• multiple correspondence analysis to map work
• rganisation practices in a two dimensional space

(associations between variables are uncovered by calculating the chi- square distance between different categories of the variables and between the individuals).

• cluster

analysis to identify different work

• rganisation

systems: we classify individuals/jobs into clusters of work organization

(we use Ward’s hierarchical clustering method and create four dummies with 1 if the worker has been classified within one cluster of work

• rganization systems and 0, otherwise)

SLIDE 28

Work Organisation Variables

Work practices are rarely used singularly, firms organise work around bundles of work responsibilities and tasks. We use cluster analysis to construct our work organization systems (4), from a set of 17 indicators describing single work practices:

1) Does your job involve doing all or part of your work in a team? 2) Does your job involve rotating tasks between yourself and colleagues? 3) Does your main paid job involve, or not ...? a) meeting precise quality standards; b) assessing yourself the quality of your own work; c) solving unforeseen problems on your own ; d) complex tasks; e) learning new things; 4) Are you able, or not, to choose or change...? a) your order of tasks; b) your methods of work; c) your speed or rate of work; 5) Is your pace of work dependent, or not, on...? a) the work done by colleagues; b) numerical production targets or performance targets; c) automatic speed of a machine or movement of a product; d) the direct control of your boss; 6) Does your main paid job involve, monotonous tasks? 7) Does your job involve short repetitive tasks of less than...? a)1 minute; b)10 minutes.

SLIDE 29

Figure 1: Multiple correspondence Analysis and identification of work organisation practices

SLIDE 30

Work Organisation systems

• Traditional
• simple organisational structure, low formalized work (informal and non-

codified), low work intensity and informal communication

• Taylorism
• high formalization of work, high level of managerial control over

employee work; constraints on the pace of work, repetitiveness and monotony of tasks

• Lean Production
• innovation skills, formalized work organization, strict monitoring, just-in-

time production, self-assessment of quality, quality norms, demand-driven constraints on work pace

• Discretionary learning
• autonomy, teamwork, rotating tasks and multi-skilling processes, could

enrich the working environment

SLIDE 31

Table 1: Work Organisation Cluster (Percent of employees by work organisation cluster)

Discretionary learning Lean Production Taylorist Traditional Mean

Team work

0.4368

0.8503

0.7628 0.2903 0.5967

Job rotation

0.3107

0.7150

0.5929 0.2374 0.4725

Discretion in fixing work methods

0.9144

0.8523

0.1494

0.2956 0.6375

Discretion in setting work pace

0.8954

0.8249

0.2275

0.3878 0.6546

Automatic constraints on work pace

0.0243

0.2780

0.5697

0.0941 0.2177

Norm-based constraints on work pace

0.1693

0.5932

0.6820

0.1339 0.3883

Hierarchical constraints on work pace

0.1671

0.4309

0.6983

0.3402 0.3792

Horizontal constraints on work pace

0.2391

0.7217 0.7497

0.2292 0.4797

Repetitiveness of tasks (every 1 min.)

0.1523

0.3420 0.5239 0.2663 0.3009

Monotony of tasks

0.2284

0.4478

0.7001 0.4906

0.4317

Responsibility for quality control

0.7812

0.9313

0.7268 0.3222 0.7334

Quality norms

0.6198

0.9363

0.9092 0.4062 0.7318

Complexity of tasks

0.6220

0.8064

0.4417

0.1602

0.5595

Learning new things in work

0.7970 0.9347

0.5693

0.2821

0.7020

Problem solving activities

0.9402 0.9532

0.6706 0.4449 0.8035

Discretion in fixing work order

0.9006

0.8354

0.1297

0.2824 0.6218

Demand constraints on work pace

0.7303 0.8127 0.5999

0.5863

0.7039

N° observations 4931 4442 2884 2666 14,923

SLIDE 32

Table 2: Descriptive statistics

% of workers reporting PH or MH

Discretionary learning Lean Production Taylorist Traditional Mean Physical health

0,0852 0.1540 0,1806 0,0759

0.1223 Skin problems 0,0424 0,0858 0,0977 0,0401 0.0655 Respiratory difficulties 0,0217 0,0490 0,0747 0,0243 0.0405 Stomach ache 0,0320 0,0563 0,0615 0,0293 0.0444 Hearth disease 0,0070 0,0133 0,0125 0,0075 0.0100 Mental health

0,2446 0,3262 0.3280 0,1951

0.2758 Stress 0,2906 0,2906 0,2810 0,1693 0.2418 Sleeping problems 0,0583 0,0942 0,1042 0,0415 0.0747 Anxiety 0,0501 0,0745 0,0923 0,0437 0.0643 Irritability 0,0726 0,1228 0,1292 0,0623 0.0965

SLIDE 33

Empirical Strategy

• We relate indicators of work-related health to a vector of work
• rganisation systems (WO), individual and job characteristics and

working conditions. We fit a probit model (and report marginal effects).

• The specification used is:

𝑸𝒔𝒑𝒄 𝑰𝒋𝒌 = ∅ 𝜷 + 𝒀𝒋𝒌

′ 𝜸 + 𝑿𝑷𝒋𝒌𝝒 + 𝜻𝒋𝒌

[1] where the dependent variable is our health indicator (Hij) either for mental or physical health for individual i, in country j.

• 𝑿𝑷𝒋𝒌 is the vector of work organisation systems.
• X is a vector of demographic (gender and age classes), job

characteristics (industry, occupation and firm size) and working conditions (working at Saturday, at Sunday, at night, at evening, more than 10 hours per day, at high speed, to tight deadlines)

• All specifications include country/time fixed effects, while εit is the error

term.

SLIDE 34

Selection

• One well known problem in the literature is

the "healthy worker effect" whereby healthy workers (w.r.t. unhealthy workers) are:

(i) more likely to be employed (ii) more likely to be in firms/jobs with more demanding work organisation systems.

• This is likely to induce “negative selection”

and bias our results downward (i.e. towards no effect)

– Note that the opposite hypothesis of “positive selection” would require less healthy workers to choose more demanding jobs which seems rather unlikely.

SLIDE 35

Tab 3: The impact of work organization systems on physical health

Physical problem Skin problems Respiratory difficulties Stomach ache Hearth disease Lean production

0.0684***

(0.0100)

0.0353***

(0.0072)

0.0207***

(0.0053)

0.0196***

(0.0054) 0.0020 (0.0019) Taylorism

0.0681***

(0.0110)

0.0310***

(0.0078)

0.0255***

(0.0063)

0.0226***

(0.0063) 0.0013 (0.0020) Discretionary learning 0.0275*** (0.0091) 0.0118* (0.0063) 0.0069 (0.0047) 0.0041 (0.0046)

• 0.0009

(0.0016) Personal characteristics YES YES YES YES YES Job characteristics YES YES YES YES YES Working conditions YES YES YES YES YES Country dummies YES YES YES YES YES Year dummy YES YES YES YES YES N°obs 14923 14923 14886 14923 14724 R-Squared 0.097 0.1026 0.1232 0.1152 0.1044

SLIDE 36

Tab 3: The impact of work organization systems on mental health

Mental problem Stress Sleeping problems Anxiety Irritability Lean production

0.0858***

(0.0132) 0.0784*** (0.0126) 0.0295*** (0.0074) 0.0258*** (0.0061) 0.0367*** (0.0087) Taylorism

0.0914***

(0.0144) 0.0774*** (0.0138) 0.0420*** (0.0089) 0.0376*** (0.0075) 0.0488*** (0.0100) Discretionary learning

0.0416***

(0.0125) 0.0353*** (0.0119) 0.0122* (0.0068) 0.0129** (0.0055) 0.0069 (0.0078) Personal characteristics YES YES YES YES YES Job characteristics YES YES YES YES YES Working conditions YES YES YES YES YES Country dummies YES YES YES YES YES Year dummy YES YES YES YES YES N°obs 14923 14923 14886 14923 14886 R-Squared 0.1056 0.1015 0.1254 0.1605 0.0793

SLIDE 37

Heterogeneity

• Personal characteristics:

– Larger effects of WO on male and younger individuals – Elders have higher % of health problems but the lowest effects of WO

• Job attributes:

– Larger effects of WO on physical health in manufacturing sector; no differences with service sector in terms of mental health – Employees in larger enterprise (+50 employees) have higher probabilities of physical and mental problems and WOs matter more wtr to those employed in smaller firm

• Working conditions:

– More detrimental working conditions (working for more than 10 hours per day, at high speed and to tight deadlines) lead to more health problems and to larger differences between WO – Pay for performance mechanism augments the probability of health problems but differences across WO are lower (and insignificant in terms of mental health)

SLIDE 38

Counterfactual Evidence

Figure 2: Kernel density by work organisation practices … moving from ‘Lean production’ to ‘Discretionary learning’

Lean production Discretionary learning Lean production Discretionary learning

SLIDE 39

Conclusions

• Recalling the main research question:

When the technological progress increases workload demands with a larger use of HIM practices, which work organization regime could help to minimize health problems?

• Traditional work organization could not be a valid alternative (only

18% of employees in 2000-2005 EWCS => low-skilled jobs with elementary tasks)

SLIDE 40

Conclusions

• Following Karasek, the answer is more control: ‘discretion’ and

‘autonomy’, typical of Discretionary learning work organization, lead to a lower probability

• f

reporting health problems (w.r.t. Lean Production and Taylorism).

• Consistent with evidence from previous studies, highly formalised

repetitive and monotonous jobs, are associated with worse health

• utcomes. Due to selection bias this is likely to be an underestimate of

the true effect.

• The implications for management is that work organisation system

characterised by more demanding tasks should encompass higher autonomy and discretion in work tasks (such as in ‘Discretionary learning’).

SLIDE 41

Thank you

gabriele.mazzolini@unimib.it

SLIDE 42

Heterogeneity – Personal Characteristics

Male Female Less than 30y 30-45y More than 45y Expected probability (Traditional) 0.0948 0.0654 0.0661 0.0876 0.0929 Lean production 0.0817*** 0.0524*** 0.0817*** 0.0752*** 0.0439** (0.0148) (0.0131) (0.0181) (0.0155) (0.0184) Taylorism 0.0858*** 0.0503*** 0.0772*** 0.0715*** 0.0490** (0.0168) (0.0139) (0.0190) (0.0174) (0.0207) Discretionary learning 0.0345** 0.0199* 0.0336** 0.0242* 0.0229 (0.0146) (0.0103) (0.0166) (0.0141) (0.0163) Observations 8,449 6,474 4,208 6,828 3,881 Pseudo R-squared 0.106 0.106 0.106 0.106 0.106 Male Female Less than 30y 30-45y More than 45y Expected probability (Traditional) 0.2432 0.2028 0.1919 0.2317 0.2500 Lean production 0.0840*** 0.0907*** 0.0863*** 0.1005*** 0.0627** (0.0185) (0.0193) (0.0226) (0.0209) (0.0252) Taylorism 0.0936*** 0.0941*** 0.0862*** 0.1281*** 0.0436 (0.0202) (0.0206) (0.0237) (0.0233) (0.0275) Discretionary learning 0.0410** 0.0427*** 0.0251 0.0806***

• 0.0052

(0.0185) (0.0165) (0.0216) (0.0203) (0.0224) Observations 8,449 6,474 4,208 6,828 3,887 Pseudo R-squared 0.106 0.106 0.106 0.106 0.106

Physical health Mental health

SLIDE 43

Heterogeneity – Job attributes

Physical health Mental health

Managers White workers Blue collars Manufac Service sector Above 50 workers Below 50 workers Exp.% (Traditional) 0.0451 0.0536 0.1272 0.1013 0.0712 0.0795 0.0838 Lean production 0.0697*** 0.0526*** 0.0868*** 0.0922*** 0.0543*** 0.0765*** 0.0644*** (0.0244) (0.0134) (0.0182) (0.0210) (0.0107) (0.0186) (0.0118) Taylorism 0.1174*** 0.0373** 0.0844*** 0.0991*** 0.0499*** 0.0900*** 0.0560*** (0.0434) (0.0151) (0.0171) (0.0211) (0.0128) (0.0207) (0.0130) Discretionary learning 0.0321 0.0221** 0.0341* 0.0371* 0.0223** 0.0394** 0.0231** (0.0211) (0.0106) (0.0180) (0.0217) (0.0089) (0.0183) (0.0102) Observations 3,765 5,046 6,112 5,453 9,461 5,447 9,469 Pseudo R-squared 0.109 0.109 0.109 0.109 0.109 0.109 0.109 Managers White workers Blue collars Manufac Service sector Above 50 workers Below 50 workers Exp.% (Traditional) 0.2652 0.2132 0.2205 0.2220 0.2275 0.2461 0.2140 Lean production 0.0783** 0.0614*** 0.1026*** 0.0875*** 0.0866*** 0.1001*** 0.0779*** (0.0356) (0.0201) (0.0208) (0.0246) (0.0159) (0.0246) (0.0155) Taylorism 0.1212** 0.0894*** 0.0888*** 0.0840*** 0.1041*** 0.1181*** 0.0742*** (0.0472) (0.0240) (0.0197) (0.0246) (0.0186) (0.0263) (0.0171) Discretionary learning 0.0110 0.0413** 0.0531** 0.0565** 0.0365** 0.0516** 0.0372*** (0.0345) (0.0178) (0.0207) (0.0257) (0.0142) (0.0248) (0.0141) Observations 3,765 5,046 6,112 5,462 9,461 5,447 9,469 Pseudo R-squared 0.109 0.109 0.109 0.109 0.109 0.109 0.109

SLIDE 44

Heterogeneity – Working conditions

All Pay for performance More than 10 hours per day Working at high speed Working to tight deadlines Exp.% (Traditional) 0.0829 0.1135 0.1007 0.1092 0.1149 Lean production 0.0684*** 0.0600** 0.0922*** 0.0800*** 0.0634*** (0.0100) (0.0286) (0.0216) (0.0161) (0.0161) Taylorism 0.0681*** 0.0775** 0.1011*** 0.0760*** 0.0677*** (0.0110) (0.0354) (0.0274) (0.0168) (0.0174) Discretionary learning 0.0275*** 0.0168 0.0412* 0.0297* 0.0236 (0.0091) (0.0285) (0.0217) (0.0163) (0.0170) Observations 14,923 2,742 4,781 7,601 7,565 Pseudo R-squared 0.0800 0.0800 0.0800 0.0800 0.0800 All Pay for performance More than 10 hours per day Working at high speed Working to tight deadlines Exp.% (Traditional) 0.2260 0.3224 0.3086 0.2894 0.3011 Lean production 0.0858*** 0.0190 0.0998*** 0.0920*** 0.0775*** (0.0132) (0.0371) (0.0279) (0.0202) (0.0210) Taylorism 0.0914*** 0.0552 0.1071*** 0.0941*** 0.0880*** (0.0144) (0.0423) (0.0320) (0.0209) (0.0221) Discretionary learning 0.0416***

• 0.0211

0.0509* 0.0401* 0.0274 (0.0125) (0.0368) (0.0284) (0.0206) (0.0219) Observations 14,923 2,742 4,781 7,601 7,565 Pseudo R-squared 0.0800 0.0800 0.0800 0.0800 0.0800

Physical health Mental health