The value of reduced risk of injury and deaths in Pakistan--using actual and perceived risk estimates.

Rafiq, Muhammad ; Shah, Mir Kalan

This study has been designed to obtain the statistical value of life and health in Pakistan by examining the compensating wage differentials among the blue collar workers of the manufacturing sector in Lahore. So far, no estimates based on compensated wage models or contingent valuation method are available for the country. Our results are based on the oneto-one interviews of 680 workers. We have estimated at the VSL and the VSI based on actual and perceived measures. The study estimates the Value of Statistical Life (VSL) to be between $ 122,047 (10.4 million PKR) and $435,294 (37 million PKR) per statistical life. Moreover, the Value of Statistical Injury (VSI) is within a range of $417 (35,445 PKR) and $1654 (140,590) per statistical injury. These values are low as compared to the values of developed countries; however, our results are akin to the results of many studies conducted in several developing countries including India, South Korea, and Mexico. The variations in the results are due to the use of different risk measures, that is, actual and professed or perceived risk measures in alternative regression equations. The regression models are fully robust and do not suffer from any major econometric problem. The results of the study will facilitate different public and private sector agencies for a better approximation of the benefits of pollution reduction and other safety measures such as traffic satiety and medical intercessions. It may also encourage further research in this area.

JEL classification: J170, J010, J300, Q510

Keywords: Compensating Wage Differentials, Lahore, Likert Scale

1. INTRODUCTION

Different safety measures adopted by governments across the globe require the estimates of willingness to pay of the people to swap wealth for a reduction in the probability of death and injury. The approximation of these trade-offs are employed in assessing the cost-benefit analysis of environmental issues, public safety measures on highways and roads, medical treatments, and many other areas. Economists term a tradeoff between money and fatality risks as the Value of a Statistical Life (VSL).

The Value of Statistical Life and Limb is generally predicted using one of the three main approaches. The first is by the compensating wage differentials that workers must be paid to take riskier jobs [Viscusi and Aldy (2003)]. The second approach examines other behaviours where people weigh costs against risks [Blomquist (2004)] and the third is through contingent valuation surveys where respondents report their willingness to pay (WTP) to obtain a specified reduction in mortality risks. The VSL is then obtained by dividing the WTP by the risk reduction being valued [Alberini (2005)].

However, most of these studies are conducted in developed countries and previously no such estimates based on willingness to pay (WTP) studies were available for Pakistan. A recent World Bank publication (1) had disclosed that the annual health effect of ambient air pollution in Pakistan includes 22,000 premature deaths among adults and 700 deaths among children under five. The total health cost of air pollution is estimated to be between .62 billion PKR to Rs 65 billion PKR or approximately one percent of GDP. It places the implied VSL figures to be in the range of 58 billion to Rs 61 billion PKR or less than three million per statistical life.

Nevertheless, these estimates are less than many regional and international studies. (2) Besides this, these estimates are based on extrapolated values from other countries, on cost of illness approach, and human capital approaches in the absence of true willingness to pay (WTP) estimates for the country. (3) Economists term such estimates as a lower bound of the premature mortality and morbidity. The absence of true estimates of VSL poses a serious problem for the policy maker in the cost-benefit analysis of different policy options.

We estimate the value of statistical life and injury in Pakistan based on compensating wage differential among the blue-collar male workers of the manufacturing sector in Lahore. We estimate the wage-risk tradeoff based on 2-digit industry level, as well as perceived measure of risk. Perceived risks are more plausible as they reflect job and work specific risks rather than industry aggregates which simply signal same level of risks for all occupations and work in a specific industrial classification. However. workers are not typically used to compute risks, this might overestimate the results. (4) To circumvent this problem we introduce two variants of the perceived fatal risk.

This is the first study of its kind in Pakistan. The results of the study shall help different agencies and research bodies for the evaluation of different safety programmes. The study will also be a springboard for further exploration and research in this area.

2. THEORETICAL IDEAS

Workers while considering the job characteristics examine many pecuniary and non-pecuniary characteristics of work, such as wages, work time career path, ease and hardship of work, pension and benefits and risk of life and health. Nonetheless, as noted by Viscusi (1978a, 1978b) that job safety is expected to be one of the most important characteristics. The theory of compensating wage differentials postulates that if a job is more riskier than the other jobs and this is known to the workers, then there must be some other more valued characteristics of that job as a compensation, but if the non-monetary aspects of all the others job are the same, then the compensation should be in the from the higher wages.

The theory was originally formed by Adam Smith who explicated that "'The wages of labourers vary with the ease or hardship, the cleanliness or dirtiness, the honorableness or dishonorableness of the employment." Economists have developed statistical models to realise the difference in workers' productivity and different component of job by unraveling wage-risk trade-off from other factors affecting wages. Griliches (1971), Rosen (1974, 1986), and Thaler and Rosen (1975) have reorganised this concept. The critique has been termed as the Hedonic (quality adjusted) Wage Model which tries to determine the variability in wages pertaining to different factors including job related fatal and non-fatal risks.

While considering the Hedonic Wage Model, the demand for labour is a decreasing function of the cost of employing labourers. These costs include wage, compensation, training and development, rest days, provision of safety measures, etc. Firms are willing to pay less to their workers as the cost of safety for a given level of profit increases. Given the wage risk offers, workers choose a wage-risk combination in the market offering highest wages. The supply of labour is fractionally influenced by their wage, risk preferences, besides numerous pecuniary and non pecuniary job characteristics.

The hedonic wage model can be explained with state-dependent utility functions. Let U(w) represent the utility of a worker in good health earning wage w and let (w) represent the utility of an injured worker at wage w. More routinely workers' are paid compensation for an injury depending upon wage one was receiving. Suppose that the compensation received by the worker and its association with the wage is symbolised by the functional form of V(w), and beside this it is also supposed, beside this it is also supposed that workers favour healthy state over an injured one, that is, U (w) > V(w). Moreover, the marginal utility of income is positive. Symbolically, U'(w) > 0 V'(w) > 0. Let p be the probability of risky event. Labours select the wage-risk deal from the available alternatives. Then the expected utility of the worker can be expressed as:

Z = (1 - p) U(w) + pV(w) (1)

And the wage-risk swapping can be expressed as:

dw/dp = -Zp / Zw = U (w) - V (w)/(1 -p) U'(w) + p V'(w) > 0 (2)

Therefore, wage must increase with the increase in the degree of risk. As a result the wage-risk swap is equated to the differentiation in the utility levels of the two states by the expected marginal utility of income. We need the observed market data to study equality between these two, and for many workers, observations of a range of workers are the combination of workers' wage and risk trade-offs. Hedonic wage models trace these loci of point by workers which is determined by the demand and supply in the market. Precisely, the coefficients match to the employee's marginal willingness to accept risk, on the other hand his demand for more safety and the firm's incremental cost for the provision of increased safety demand plus the decrease in the marginal cost faced by the firm owing to more risk faced by the worker. (5)

Data and Variables

For estimation of the hedonic wage equation, take home hourly wages have been used as a dependent variable. This was obtained directly from the respondents. (6) The independent variables include risk variables such as annual average fatalities per 10,000, nonfatal accident per 100 workers, human capital variables such as age, education, experience, and job characteristics such as type of permanent or temporary jobs, job related trainings compensation provided by the company in case of industrial accident etc. industrial dummy variables to obtain difference in the wage among different industrial classifications, and professional dummy variables to control for differences in the wages among different professions such as supervisor, motor operators, electricians and foreman etc.

The data pertaining to worker's fatal accident for the year 2006-2007 was compiled from the records of the Punjab Employees Social Security Institute (PESSI). The institute does not regularly publish these incidents, so the record had to be compiled manually by looking into the registers which were maintained in their main and sub offices across different parts of Lahore. (7) Ironically, even the Federal Bureau of Statistics and Punjab Bureau of Statistics do not publishes details of industrial fatal accidents.

The data pertaining to non-fatal accidents pet 100 workers was compiled from the data set of the Labour Force Survey (LFS) (8) (2006). Non-fatal risks have also been used as one of the explanatory variable in this study. However, we have employed two different types of non-fatal risks. Both have been obtained from the LFS. (9) This has been done to analyse the difference in the respective Values of Life and limb. The two measures of injuries have been used in separate equations. One of such measures is the Punjab nonfatal industrial accidents among the manufacturing sector workers for the year 2006, whereas the other is the, country wise industrial non-fatal accidents for the same year.

But these fatal and non-fatal risk data are two digit (10) industrial risk averages. However, perceived fatal and non-fatal risks were elicited using Likert scales. Separate scales were used for the risk of death and the risk of injuries. These scale ranged from 1-5, where I represent minimal and 5 a maximal risk of receiving fatal and non-fatal accidents. (11)

However, following the work of Hammitt and Ibarraran (2006) and others, beside these two measures of perceived risks, another measure was also developed for obtaining the perceived fatal risk. A scale which ranged from 0-10 out of 10,000 was used. (12) As an example, 0/10,000 chances means no chance of risk and 10/10,000 refers to .001 chances of receiving job related fatal accidents. Verbal analogies were used in order to help the respondent answer the question.

We tried two analogies including an explanation such as numbers of hours in fourteen month which are approximately 10,000 and secondly a scenario describing the chances of receiving job related fatal injuries out of 10,000 of people doing the same job as you are doing. We only used second analogy when we realised that the first one is not helping them answer the question and the majority of them could only understand it with the second analogy.

Sampling and Primary Data Collection

Multi stage sampling technique was used for data collection. At the outset, Lahore was selected as the study area because it is the second largest industrial city and is also a nearest study destination. For the interview, the blue collar male workers of the manufacturing sector who had also served in Lahore for at least a year were selected. (13) The survey was also limited to the workers of the factories registered under industrial act 1934. By this means the survey was confined to the formal sector. It was also important to confine the survey to the formal sector because of the fact that the formal sector's labour market is not distorted and the wages were determined by demand and supply. (14) Further stratified random sampling technique was adopted to draw out the representative sample. The stratification was done based on the National Industrial Codes (NIC) which has classified the industrial group in to nine industrial categories.

For determining the sample size precedent was used as many other regional and international studies have employed a sample size of more than a 1000 workers (15), hence it was taken as a precedent and the sample size was set down as 1000 blue collar male workers. Interestingly, the sample size also turned out to be ten percent of the manufacturing workers in Lahore.

The factories and respondents were randomly picked up; as an example any seven to ten workers were interviewed from the concerned industrial classification. However the number of industrial unit per industrial classification and the number of respondents per factory was based on the risk categories. The reason for including more workers and factories from high risk categories was to allow the variation in the data. The risk categories were obtained from the Labour Force Survey for the year 2006. (16)

A survey was designed to collect data from the workers of the manufacturing sector. In person interviews were conducted from the blue-collar male workers. The questionnaire was pretested in a pilot study of fifty workers. The results of the pilot study were used to strategise the data collection procedure. During the said study it was observed that the industrialists were hesitant to allow their workers to be interviewed. Beside this, it was also observed on few occasions that the workers were instructed not to answer few questions. Therefore, for the final survey a three prong approach was adopted for interviewing the respondents, Firstly, by contacting the employers, secondly, by visiting the cafeterias inside industrial zones during lunch or tea time, and a third, by going to the residential compounds/villages on off days.

The survey started in April 2009, and was extended to all the parts of Lahore including industrial zones, housing colonies and the villages on the peripheries. The main industrial zones are situated on Ferozpur road, Multan roads, Quaid-i-Azam industrial estate, Sundas industrial estate, industries situated on Rai Wind road. Moreover, approximately, fifty five villages on the fringes of Lahore were also expedited for interviewing the workers.

But, due to deteriorating law and order situation the survey was discontinued in October, 2009. Because of this reason, six hundred and eighty respondents were interviewed which is still more than the required number, as per the sampling formula. Table 2 shows the actual number of respondent as against the target in each industrial group.

Econometric Model

The data is analysed through the estimation of hedonic wage equations by regressing log of hourly wages on human capital variables, industrial dummy variables and occupational dummies. The hedonic wage equation is given as follows:

Ln[W.sub.i] = [beta] + [H.sub.i] [[beta].sub.1] + [[chi].sub.i] [[beta].sub.2] + [p.sub.i] [[beta].sub.3] + [q.sub.i][[beta].sub.4] + [[epsilon].sub.i] (4)

Where, LnWi is the worker i's hourly wage rate in logarithmic term, [alpha] is a constant term, H is a vector of personal characteristic variables for the worker i. This include education measured as years of education, age and experience, X is a vector of job characteristic which comprises, training and compensation variables, six industries dummy, three profession dummy variables, a variable to denote whether the job is permanent or temporary. Di is the fatality risk associated with worker i's job per 10,000 workers, and Ni is the nonfatal injury risk associated with worker i's job per 100 workers, and [[epsilon].sub.i] is the random error.

The dependent variable has been measured as hourly wage rates; evidently many other studies have also used hourly wage rates. However, the choice of the functional form is an unanswered question. Different researchers have used either linear or log-linear form. Subsequent upon the Meta analysis of Viscusi and Aldy (2003), present study has made use of Box-Cox transformation to decide about the dependent variable. We estimated both the linear form and the log form of wages in the resilient Box-Cox transformation, yet it reinforced both the functional form when a log form was used and it supported none when linear form was employed. (17)

Value of Statistical Life and Value of Statistical Injury were computed using the following equations:

VSL = [beta]'3 x [W.sup.-1] x 2000 x 10000 & VSI = [beta]'4 x [W.sup.-] x 2000 x 100 (5)

Where,

[beta]'s are the respective risk coefficients, [W.sup.-] is the mean hourly wage rate which is multiplied with the 2000 (18) annual hours of work to annualise the Value and is multiplied with the scale of the variable which is per 10,000 workers for the fatality risk variables and per 100 worker for the non-fatal risk variable.

RESULTS AND DISCUSSION

The descriptive statistics along with the definition of the variables which have been used for the present analysis are in Table 3. The average hourly wage rate in log form is 3.705 (anti-log = 42PKR (19)). Average education is six years of schooling and average age is 27 years. Average experience in the present occupation is 5 years.

The 2-digit industry level fatality rate and the perceived fatality rate are almost similar with a slight variation that is 1.17 and 1.36 per 10,000 per annum. The professed fatality and non-fatality statistics measured on Likert scale reflect mean risks as perceived by workers is below average level of risk (mean risk = 3). The industry level injury averages for both Pakistan as a whole and Punjab-wise are modestly close that is, 4.14 and 3.9 per 100 workers per year respectively.

The estimation results of the alternative hedonic wage models are presented in Table 4. Column 1 and 2 of the Table show the regression results based on 2- digit industry level fatal and non-fatal risk variables, whereas, column 3 and 4 are explicating the regression estimates using the perceived risk measures.

The coefficient of fatal risk in all the five models using either industry level actual risk data, or individual level perceived risk measure, is positive and statistically significant. This clearly authenticates the compensating wage differentials theory and establishes that labour markets in Pakistan do pay wage premium for higher risk. However, non-fatal risk coefficient is significant when actual risk data is used.

The coefficients of fatal and non-fatal variables and subsequently the VSL and VSI in column one, is substantially higher as compared to the estimates in column two. Both the models include the same fatal risk variables, however, the former incorporates the country level non-fatal risks statistics, whereas the latter has used province wise risk data. But in our opinion the results of both the models are not directly comparable owing to different model specification. Nonetheless, this does points out the variation in VSL and VSI to the use of different risk measures and right hand side variable. The choice of the right hand side variables is based on the Likelihood Ratio (LR) test.

Similar variations are observed when the two variant of the perceived fatal risk variable along with the same non-fatal risk data are used. The VSL in column 4 which is based on the workers' perception measured on a scale 0-10/10,000 is considerably high not only as compared to the VSL estimates from alternative perceived fatal risk estimate in column 3, but is also higher than any other model. However, the model is also differently specified. The choice of the covariates in the entire estimated regression models is based on the LR test.

However, to check the robustness of our results, we have also estimated a model which includes all the industrial dummies except one. Column 5 is showing the results of such a regression. The regression model includes objective measure of fatal risk variable, but it does not include the injury variable. The coefficient of the risk variable is the same as in column 1.

The coefficients of the human capital variables are not sensitive to the choice of the other explanatory variables in the model. Both the age and education are showing positive and significant relationship with the hourly wage in all the estimated regression models, however, the result of the work experience is insignificant in all the estimated regression models. The results of the professional dummy variables are also robust and are showing little sign of variations. The outcome of these two variables shows that supervisors and foreman on the average earn 36 percent and 41 percent more than all other professional categories.

One of the industrial dummy variables, that is textile, has shown consistent results and it shows evidence of higher earnings of this group as compared to the base category. The results of other industrial classification are mixed and the coefficients are also changing signs in different specifications. This may be due to the multicollinerity problem, however, the results of the partial correlation do not show any sign of it.

Evidently, within one of the specified model, the coefficient results elucidate that workers of permanent status earns more on the average, whereas, workers who had received compensation for job related non-fatal accidents in the past receive low wage. Both the coefficients are statistically significant.

We have confirmed the structural stability of our regression models by restricting the estimations to 384 (20) respondents as was set by the sampling formula. The results are quiet robust and there has been no considerable changes in the results of the estimated coefficient.

The Value of Statistical Life and Value of Statistical Injury are shown in the Table 4. VSL based on actual risks is between $122,047 and $313,411. Whereas, VSL based on perceived risks is between $122,811 and $435,294. The VSL based on actual risk in column 2 and that in column 3 based on perceived risks are akin. The Value of Statistical Injury based on actual risks is within a range of $417 and $1654.

These values are smaller as compared to the VSL of many developed countries which is in the range of $4 million and $9 million, however our results are comparable with the estimates of many developing countries, including Mexico, India, South Korea, and Hong Kong. (21) Table 5 shows the comparison of the VSL and VSI for the developing countries.

Calculating VSL for Pakistan Based on Prediction Equation

In order to reinforce the validity of our estimates, we have also computed the Value of Statistical Life for Pakistan based on the Bowland and Beghin (2001) prediction equation which can be used to estimate the VSL for the developing countries. The equation is based on the Meta Analysis of the industrialised countries and it takes in to account the difference in risk, human capital and income between the developed and developing countries. The income elasticity estimated by the ranges from 1.52 to 2.269. (23) However, we have used the income elasticties estimated by different studies to compute Value of Life for Pakistan. Table 6 (24) present the VSL based on the prediction equation. The equation provides us a range of VSL from $0.17 million to $1.2 million, nevertheless, Miller's estimated range of elasticities gives a close approximation of our reported results.

CONCLUSION

This is the first study of its kind in Pakistan. Previously there have been no estimates available for the country based on either compensated wage models or contingent valuation method. Subsequent upon the results of the estimations, the study concludes that the Compensating-wage differential does exists in the formal private sector in Pakistan and the market does compensate the workers for taking risk. Moreover, since these compensating differentials are the consequence of labour demand and supply, therefore the hypothesis that the workers are rational and they do consider risk while accepting jobs, is therefore fully validated. The study has estimated the Value of Statistical Life (VSL) to be between $ 122,047 and $435,294 per statistical life. Moreover, the Value of Statistical Injury (VSI) is within a range of $417 and $1654 per statistical injury. The variations in the results are due to the use of different risk measures, that is, actual and professed or perceived risk measures in alternative regression equations. The regression models are fully robust and do not suffer from any econometric problem. The usual econometric problems, such as Hetroscedsticity, and specifications biases have been fully taken care off. In addition to this it is also concluded that the models are structurally stable model and the results based on a sample size of 384 respondents and that of 680 respondents do not vary dramatically. These values are robust and can be used for the cost-benefit analysis (CBA) of the safety projects in Pakistan pertaining to abatement of pollution, medical intercession and highway safety measure etc. It can be also be used for settling claims on insurance companies and other court settlement cases etc. Moreover, in the context of ongoing war on terrorism, policy maker can use it for evaluating the impact assessment of different policy options. The results of the study provide a breeding ground for supplementary exploration and research in this area.

REFERENCES

Alberini, A., M. Scasny, and M. Braun Kohlova (2005) The Value of Statistical Life in Czech Republic. Evidence from a Contingent Valuation Study. Paper submitted for presentation at the EAERE annual meeting.

Blomquist, Glenn C. (2004) Self-protection and Averting Behaviour, Values of Statistical Lives, and Benefit Cost Analysis of Environmental Policy. Review of Economics of the Household 2, 69-110.

Bowland, B. J. and J. C. Beghin (2001) Robut Estimates of Value of a Statistical Life for Developing Economies. Journal of Policy Modelling 23, 385-396.

Gerking, S., M. De Haan, and W. Schulze (1988) The Marginal Value of Job Safety: A Contingent Valuation Study. Journal of Risk and Uncertainty 1:2, 185-200.

Griliches, Z. (ed.) (1971) Price Indexes and Quality Change. Cambridge: Harvard University Press.

Gunatilake, H. M. (2003) Environmental Valuation: Theory and Application. Postgraduate Institute of Agriculture, University of Pradeniya.

Hamermesh, D. S. (1999) Changing Inequality in Markets for Workplace Amenities. Quarterly Journal of Economics 114:4, 1085-1123.

Hammitt. K. James and M. E. Ibarrara'n (2006) The Economics Value of Fatal and Nonfatal Risks in Mexico City Using Actuarial and Perceived Risk Estimates Health Econ. 15: 1329-1335, published online in Wiley InterScience (www.interscience. wiley.com). DOI: 10.1002/hec.1137

Hersch, J. and W. K. Viscusi (1990) Cigarette Smoking, Seatbelt Use, and Differences in Wage-Risk Tradeoffs. Journal of Human Resources 25:2, 202-227.

Hintermann, Beat, Alberini Anna and Markandya Anil (2006) Estimating the Value of Safety with Labour Market Data: Are the Results Trustworthy? The Fondazione Eni Enrico Mattei Note di Lavoro Series Index: http://www.feem.it

Johannesson, Magnus, Per-Olov Johansson, and Karl-Gustav Lofgren (1997) On the Value of Changes in Life Expectancy: Blips versus Parametric Changes. Journal of Risk and Uncertainty 15, 221-239.

Jones-Lee, M. W. (1976) The Value of Life: An Economic Analysis. London: Martin Robertson.

Kim, S. W. and P. V. Fishback (1999) The Impact of Institutional Change on Compensating Wage Differentials for Accident Risk: South Korea, 1984-1990. Journal of Risk and Uncertainty 18:3, 231-248.

Krupnick, A., A. Alberini, M. Cropper, N. Simon, B. O'Brien, R. Goeree, and M. Heintzelman (2002) Age, Health, and the Willingness to Pay for Mortality Risk Reductions: A Contingent Valuation Survey of Ontario Residents. Journal of Risk and Uncertainty 24, 161-184.

Leigh, J. P. and R. N. Folsom (1984) Estimates of the Value of Accident Avoidance at the Job Depend on the Concavity of the Equalising Differences Curve. Quarterly Review of Economics and Business 24:1, 56-66.

Liu, J. T., J. Hammitt, and J. L. Liu (1997) Estimating Hedonic Wage Function and Value of Life in Developing Country. Economics Letters 57, 353-358.

Liu, J. T. and J. K. Hammitt (1999) Perceived Risk and Value of Workplace Safety in a Developing Country. Journal of Risk Research 2:3, 263-275.

Madheswaran, S. (2004) Measuring the Value of Life and Limb: Estimating Compensating Wage Differentials among Workers in Chennai and Mumbai. (SANDEE Paper No. 9-04).

Moore, M. J. and W. K. Viscusi (1998a) Doubling the Estimated Value of Life: Using New Occupational Fatality Data. Journal of Policy Analysis and Management 7:3, 476-490.

Olson, C. A. (1981) An Analysis of Wage Differentials Received by Workers on Dangerous Jobs. Journal of Human Resources 16:2, 167-185.

Pakistan EPA/ World Bank (2006) Strategic Country Environmental Assessment: Rising to the Challenges. Draft May 2006. Available: http://www.environment.gov.pk/ NEWPDF/Pak-SCEA-May2006.pdf

Persson, U., A. Norinder. K. Halte, and K. Gralen (2001) The Value of a Statistical Life in Transport: Findings from a New Contingent Valuation Study in Sweden. Journal of Risk and Uncertainty 23:2, 121-134.

Rosen, S. (1974) Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy 82, 34-55.

Rosen, S. (1986) The Theory of Equalising Differences. In O. Ashenfelter and R. Layard (eds.) Handbook of Labour Economics. Amsterdam: North-Holland. 641-692.

Shanmugam, K. R. (1996/7) The Value of Life: Estimates from Indian Labour Market. Indian Economic Journal 44:4, 105-114.

Shanmugam, K. R. (1997) Value of Life and Injury: Estimating Using Flexible Functional Form. Indian Journal of Applied Economics 6:3, 125-136.

Shanmugam, K. R. (2000) Valuations of Life and Injury Risks. Environmental and Resource Economics 16, 379-389.

Shanmugam, K. R. (2001) Self Selection Bias in the Estimates of Compensating Differentials for Job Risks in India. Journal of Risk and Uncertainty 22:3, 263-275.

Thaler, R. and S. Rosen (1975) The Value of Saving a Life: Evidence from the Labour Market. In N. E. Terleckyj (ed.) Household Production and Consumption. New York: Columbia University Press. 265-300.

Viscusi, W. K. (1978a) Labour Market Valuations of Life and Limb: Empirical Evidence and Policy Implications. Public Policy 26:3, 359-386.

Viscusi, W. K. (1978b) Wealth Effects and Earnings Premiums for Job Hazards. Review of Economics and Statistics 60:3, 408-416.

Viscusi, W. K. (1979) Employment Hazards: An Investigation of Market Performance. Cambridge, MA: Harvard University Press.

Viscusi, W. K. (1980) Union, Labour Market Structure, and the Welfare Implications of the Quality of Work. Journal of Labour Research 1:1, 175-192.

Viscusi, W. Kip and Joseph E. Aldy (2003) The Value of a Statistical Life: A Critical Review of Market Estimates throughout the World. Journal of Risk and Uncertainty 27:1, 5-76.

World Bank (1992) World Development Report 1992. Development and the Environment. www.wikipedia.com, www.environment.gov.pk

Comments

This is an excellent study and the authors rightly claim that it is the first of its nature in Pakistan. The study tries to calculate the statistical value of life and health in Pakistan by estimating Value of Statistical Life (VSL) and Value of Statistical Injury (VSI). These values can be used, as the authors mention in the study, in the cost benefit analysis of various environmental, health, and public safety policies. However, I consider it more valuable in the sense that these estimated values can be used to calculate our human losses, both in terms of injuries and deaths, in war on terror. The estimates of losses for Pakistan in this war various between $34 billion and $60 billion, which are usually calculated on basis of non-human physical losses. Including the monetary value of human losses on the basis of these estimates will show a relatively true picture of sufferings to Pakistan. Moreover, the compensation that the government pays to the victims of terrorism ranges between 100,000 Rs to 300,000 Rs for dead and only 50,000 Rs for injured. In the light of these estimates VSL and VSI, these compensations are awfully low.

As mentioned earlier, this study is an excellent effort; nonetheless, the following queries need to be addressed.

* The authors have taken education as the years of schooling, and it has significant positive effect on the hourly wages. This means that if one year of schooling increases, the wage will also increase. However, it makes no sense that it" the year of schooling of a blue collar worker will increase from 3rd to 4th year, or from 6th to 7th year, his wage will increase. Though, this can matter if the year of schooling increase from 9th to 10th year and the worker is matriculate or higher certificate or degree holder. It may be argued that higher education will lead the worker to a higher rank job. Nevertheless, the profession dummies have already captured that effect. The problem with this measure of schooling is that it gives equal weights to each year increase. I suggest that the authors should, instead, make different categories of education and give codes to each category. This will capture the true effect of education.

* Same is the problem with the "age" variable. It is having positive effect on wage for each birthday of a worker. This may also be categorised in different age groups.

* Except for textile dummy, the coefficients of different industrial dummies are volatile both in terms of sign and significance which makes the stability doubtful. The authors should provide a satisfactory justification for this volatile behaviour.

Muhammad Nasir

Pakistan Institute of Development Economics, Islamabad.

(1) EPA/World Bank (2006).

(2) See Madheswaran for estimates of VSL in India (2004).

(3) EPA/World Bank (2006).

(4) Hammitt and Ibarraran (2006).

(5) This section is based on the meta analysis of Viscusi and Aldy (2003).

(6) The respondents had reported monthly wages which were annualised and then were divided by 2000 hours to obtain hourly wages. The 2000 hours is a standard annual work time and many studies including Viscusi and Aldy (2003) and Madesh (2004) had used similar wage estimates in their respective studies. The same is more or less true for Pakistan.

(7) We are especially thankful to Mr Safdar Raja and his team for helping us with the compilation of fatality data.

(8) I am especially thankful to Mr Masood Ashfaq and Mr Tayab at PIDE, Islamabad for helping me in obtaining the LFS data set.

(9) LFS is annually conducted by the Federal Bureau of Statistics (FBS).

(10) 2-digits refers to main industrial classifications, for example 31 represent food, beverages and tobacco industries.

(11) The questions were "please tick the appropriate box below indicating your perception of receiving a job related injury/fatality in your present job in comparison to any other job you can do.

(12) The spearman's correlation between the two perceived risk measures is found to be .51 and is statistically significant result. The relationship is not too high, but the relationship is positive and significant. This shows the consistency of the workers response.

(13) This was done to ensure that interviewee knew the labour market situation and were aware of the job related risk.

(14) There was no sample selection bias because the informal markets arc not fully functioning and the market is really distorted. Moreover, in the formal sector though there are minimum wage laws however, those are hard to implement and the role of unions is minimal.

(15) Sec Madesh (2004) and Viscusi and Aldy (2003).

(16) See annexure-3 for further details.

(17) Evidently, many other researchers, for example Moore and Viscusi (1988a), and Madeshwaran (2004) have employed the same technique. Gunatilake (2003) have also suggested making use of Box-Cox technique for selecting the functional form for such studies. The theta values = 0 was accepted when we used hrwge as dependent, however, when I used lhrwge the hypotheses that theta = 1 was accepted. It would be good to present the estimated parameters for Box-Cox transformation. That will make things easier to understand.

(18) This has been done to follow a standard practice. However, there is no change in the results if we use the log of monthly wages.

(19) This was calculated at the prevailing exchange rate which was 1 US$=85PKR.

(20) These 384 observations were randomly generated in SPSS.

(21) See Viscusi and Aldy (2003).

(22) The table has been partly developed from the study of Hammitt and Ibarraran (2006).

(23) See Brajer and Rehlnatian study "From Diye to Value of Statistical Life: A Case Study of Islamic Republic of Iran".

(24) For developing this table we have taken help from e Meta Analysis of Viscusi and Aldy (2003). USEPA and World Development Indicators (WDI).

Muhammad Rafiq <sufi92@hotmail.com> is Assistant Professor, Institute of Management Sciences, Peshawar. Mir Kalan Shah <mkshah_ids@yahoo.com> is Professor, Institute of Development Studies, KPK Agricultural University Peshawar.

Authors' Note: I fully acknowledge the financial and technical support provided by the South Asian Network for Development and Environmental Economics (SANDEE). I am highly indebted to professors Jeffery Vincent for his Valuable contribution and his keen interest in my study. I am thankful to Prof. M. N. Murty, Prof. Enamul Haque, Prof. Pranab Makapodya and anonymous reviewers for their important comments and suggestions. I am especially grateful to Dr Priya Shyamsundar for her commendable comments and constant encouragement throughout this study. I am thankful to Mr Safdar Raja (PESSI), Mr Masood Ashfaq (PIDE), Mr Saqlain Raza my survey supervisor and Qaiser for their contributions.

Table 1
Sampling Frame

Details No. of Max per
 Respondents Factory

31 Food Group 125 10
32 Textile Group 83 7
33 Wood and Furniture 125 10
34 Paper and Publishing 83 7
35 Chemical Group 83 7
36 Non Metallic 125 10
37 Metal Group 125 10
38 Fabricated Metal 125 10
39 Other 125 10

Table 2
Sample Target Versus the Actual Numbers of Respondents

NIC Type of Manufacturing Target Per Actual
 Factory Numbers

31 Manu. of food. beverages and 125 10 coax 121
 tobacco

32 Manuf. of Textile, wool and 83 7 max 82
 hosiery etc.

33 Manuf. of wood or wood 125 10 max 31
 product or furniture respondents

34 Manuf. of paper, paper prod. 83 7 max 74
 Printing publishing respondents

35 Manuf. of Chemical petroleum,
 coal rubber and plastic prod. 83 7 max 93
 respondents

36 manuf. Non-metallic product
 except petroleum and coal 125 10 max 41
 respondents

37 Basic metal industries 125 Do 91
 respondents

38 Manuf. Fabricated metal
 product machinery and 125 Do 116
 equipment respondents

39 Other manuf. Industries and 125 Do 30
 handicraft respondents

 Total Respondents 1000 680

Table 3
Variable Definitions and Descriptive Statistics

Variable Variable Definition Mean Std. Dev.

PRMNT 1 if the worker's job is permanent, 0.35 0.48
 0 otherwise

LHRWG hourly wage in PKR (in logarithm) 3.705 0.304

EDUCN years of schooling 6.037 4.129

AAAGE age of the respondent 27.38 7.983

FAMLZ family size 6.544 2.791

DEPEN No of dependents 4.46 2.275

SPEDY 1 if the worker job requires speedy 0.73 0.44
 work, 0 otherwise

EMPFM Employed family members 3.11 1.301

RGRHR Regular hours of work 8.697 1.612

EXPER experience in years 4.842 5.893

DSTNC Distance from the work place in 31.36 30.78
 minutes

UNION 1 if union member, 0 otherwise 0.0265 0.16

DCNMK 1 if the worker has to make 0.43 0.50
 decision, 0 otherwise

TRNNG 1 if the worker is provided any kind 0.84 0.36
 of training, 0 otherwise

PESFAT 2-digit fatality rate compiled from 1.17 1.27
 the office of Pmijab Employees
 Social Security Institute per 10,000
 workers

LFSPK 2-digit injury rate of Pakistan's 4.14 2.3
 manufacturing worker computed from
 the labour force survey (LFS, 2006)
 per 100 workers

LFSPN Injury rate of Punjab based 3.9 1.88
 manufacturing worker computed from
 the labour force survey (LFS, 2006)
 per 100 workers

PRFNJ Professed/perceived injuries 2.26 1.14
 proportion measured on a liken scale
 1-5 scale

PRFT1 Professed/perceived fatalities 1.27 0.68
 proportions measured on a liken
 scale 1-5

PRFT2 Professed/perceived fatalities rate 1.36 2.138
 0-10 per 10000

TOTMP Total no of employees 501 1108

LFINS 1 if the worker life is insured, 0 0.08 0.29
 otherwise

COMPS 1 if the worker is provided 0.52 0.51
 compensation by the employers, 0
 otherwise

WTHDM Wealth dummy= value of the house in 885136 1159939
 PKR

NMSTK 1 if the worker job requires no 0.15 0.37
 mistake, 0 otherwise

JBNOS 1 if the worker job is very noisy, 0 0.8 0.4
 otherwise

EXPOS 1 if the worker is exposed to smoke 0.63 0.48
 or dust, 0 otherwise

TXTDM 1 if the worker is from the Textile 0.13 0.33
 group, 0 otherwise

BSCMT 1 if the worker is from Basic metal 0.13 0.34
 group, 0 otherwise

SPORT 1 if the worker is front Spoil and 0.04 0.2
 others group, 0 otherwise

WOOD 1 if the worker is from wood and 0.04 0.2
 furniture group, 0 otherwise

FOOD 1 if the worker is from the food 0.17 0.38
 group, 0 otherwise

PAPER 1 if the worker is from the paper 0.10 0.31
 group, 0 otherwise

CHEME 1 if the worker is from the chemical 0.13 0.34
 group, 0 otherwise

FABRI 1 if the worker is front the 0.17 0.37
 fabricated metal group, 0 otherwise

DSTRT 1 if the worker is from district 0.71 0.45
 Lahore, 0 otherwise

SUPER 1 if the worker is a supervisor, 0 0.036 0.18
 otherwise

MACOP 1 if the worker is a machine 0.23 0.43
 operator, 0 otherwise

FORMN 1 if the worker is a foreman, 0 0.04 0.2
 otherwise

Table 4
Regression Results of the Alternative Heclonic Wage Equations

Variables (1) (2) (3)

PRMNT -- 0.063 *** --
 (0.02)
EDUCN 0.013 *** 0.011 *** 0.015 ***
 (0.003) (0.003) (0.003)
AAAGE 0.009 *** 0.007 *** 0.008 ***
 (0.002) (0.001) (0.002)
EXPER 0.003 0.003 0.002
 (0.003) (0.002) (0.002)
TRNNG 0.02 -- --
 (0.03)
PESFAT 0.361 *** 0.141 *** --
 (0.105) (0.03)
LFSPK 0.19 *** -- --
 (0.068)
LFSPN -- 0.054 *** --
 (0.02)
PRFNJ -- -- 0.06
 (0.08)
PRFTI -- -- 0.156 *"' *
 (0.06)
PRFT2 -- -- --

COMPS -- 0.08 *** --
 (0.02)
TXTDM 0.949 *** 0.449 *** 0.119 ***
 (0.295) (0.095) (0.044)
BSCMT -0.39 *** -- 0.165 ***
 (0.13) (0.042)
SPORT -- -- --

WOOD -- 0.062 --
 (0.064)
FOOD -- 0.112 *** --
 (0.04)
PAPER 0.11 -- 0.016
 (0.11) (0.041)
CHEME 1.069 *** 0.338 *** -0.02
 (0.37) (0.105) (0.03)
FABRI -0.185 *** 0.07 * --
 (0.067) (0.04)
NONMETL -- -- --
SUPER 0.401 *** 0.356 *** 0.366 ***
 (0.098) (0.07) (0.104)
MACOP -- -- --
FORMN 0.41 *** 0.385 *** 0.443 ***
 (0.084) (0.06) (0.08)
EXPERSQ -- -- --
[R.sup.2] 0.25 0.25 0.21
F 11.5 15.84 12.44
VSL (PKR) 26,640,000 10,374,000 11,554,000
VSL@85PKR/$ $313,411 $122,047 $135,811
VSI@85PKR/$ $1,654 $470 $523

Variables (4) (5)

PRMNT -- --

EDUCN 0.013 *** 0.01 ***
 (0.0028)
AAAGE 0.008 *** 0.03 ***
 (0.0018)
EXPER 0.003 0.004
 (0.0026)
TRNNG -- --

PESFAT -- 0.36 **

LFSPK -- --

LFSPN -- --

PRFNJ 0.049 --
 (0.0901)
PRFTI -- --

PRFT2 0.542 ** --
 0.2408
COMPS -- --

TXTDM 0.169 *** --
 0.0504
BSCMT 0.22 -I.I
 0.0558
SPORT 0.219 -1.04 *
 0.056
WOOD -- -0.006

FOOD -- -0.15 ***

PAPER 0.072 -0.9 **
 0.054
CHEME 0.052 0.02
 0.0449
FABRI 0.094 ** -0.2 ***
 0.0481
NONMETL -- -0.03
SUPER 0.369 *** 0.35 ***
 0.0981
MACOP -- -0.01
FORMN 0.427 *** 0.4 ***
 0.0834
EXPERSQ -- -0.00004
[R.sup.2] 0.22 0.24
F 11.15
VSL (PKR) 37,000,000
VSL@85PKR/$ $435,294
VSI@85PKR/$ $427

Note: The parentheses are showing robust standard errors of the
estimates except for the second model. This is due to the fact that
hetroscedsticity test for the second model was insignificant.

Table 5
Comparative Statistics of VSL and VSI of Developing Countries 22

 Average Average
 Income Fatal Risk
Study Country (2000 US $) (per 10000)

Hammitt and Ibarraran Mexico 4100 3.0
Kim and Fishback South Korea 8100 4.9
Liu, et al. Taiwan 5000-6100 2.3-3.8
Liu, et al. Taiwan 18500 5.1
Shanmugun India 780 1.0
Shanmugun India 780 1.0
Shanmugun India 780 1.0
Madesh India 780 1.13
Siebert and Wei Hong Kong 11700 1.4

 VSL VSI
Study (2000 US $) (2000 US $)

Hammitt and Ibarraran 230000-310000 3000-10.000
Kim and Fishback 800,000
Liu, et al. 200.000-900.000
Liu, et al. 700,000 50,000
Shanmugun 1,200,000-1,500.000
Shanmugun 1,000,000-1.400,000 150.000-560,000
Shanmugun 4,100.000 350.000
Madesh 305.000-318,000
Siebert and Wei 1,700.000

Table 6
VSL for Pakistan Based on Prediction Equation Using Different
Income Elasticities

 Income US GNI Pakistan
 Elasticity per Capita per Capita
Study ([alpha]) (2008) (2008)

Miller (2000) 0.85 $47,930 $950
Miller (2000) 0.96 $47,930 $950
Morzek and Taylor (2006) 0.46 $47,930 $950
Morzek and Taylor (2006) 0.49 $47,930 $950
Viscusi and Aldy (2006) 0.52 $47,930 $950
Viscusi and Aldy (2003) 0.61 $47,930 $950

 VSLpk=
 VSLus
 [(GNIpk/
 GNIus).sup.
Study US VSL [alpha]

Miller (2000) $7,400.000 $264,107
Miller (2000) $7,400.000 $171,578
Morzek and Taylor (2006) $7,400.000 $1,218,723
Morzek and Taylor (2006) $7,400.000 $1,083,474
Viscusi and Aldy (2006) $7,400.000 $963,234
Viscusi and Aldy (2003) $7,400,000 $676,819