The dynamics of moonlighting in Pakistan.
Hyder, Asma ; Ahmed, Ather Maqsood
The study explores the dynamics of moonlighting, demographics,
human capital and association of occupations between primary and
secondary job. The paper is based on cross-section data Labour Force
Survey 2006-07 and limited to male wage workers residing in urban areas.
Among two motives according to theoretical framework of moonlighting;
first, constraint on hours worked in first job and second is wage rate
is lower than the reservation wage in the primary occupation; within
limited information available on different variables our results are
skewed toward first motive and earnings from the primary occupation are
insignificant in moonlighting decision. The model specification also
attempts to correct the endogenous regressor in probit estimation. Among
moonlighters 'Professionals' and 'Technicians' are
holding their secondary job in same occupational category; apart from
these two occupational categories managers and elementary occupations
also seems popular for moonlighting.
JEL classification: J22, J24
Keywords: Moonlighting, Labour Mobility, Occupational Association
1. INTRODUCTION
There is substantial amount of literature available in Pakistan
focusing on wage level and trends in different sectors of the economy to
determine labour market outcomes and differences in the living standard
of workers. Irfan (2008) has shown that the time trend growth in wages
for the last two and half decades has been 7.7 percent in contrast to
7.2 percent in prices thereby yielding 0.7 percent growth in real wages.
This confirms that the purchasing power of consumers has remained very
low due to surge in prices of commodities of consumer basket.
Unfortunately the economy of Pakistan is facing multiple challenges,
starting from its difficulty in arresting the escalation in prices, to
achieving sustainable growth in a longer term horizon. There is
virtually no improvement in socio-demographic indicators. The labour
market dynamics are such that it is becoming increasingly difficult to
train and educate people for work and to provide them with decent
well-paying jobs. Resultantly, there is a dearth of educated and skilled
people, and the mismatch between jobs and workers has increased. In this
scenario, those who posses some sort of skills or specific education,
they try to get maximum benefit out of it, including holding multiple
jobs, even if it involves jobs in other than their own primary
occupation.
The moonlighting of holding of dual jobs has not been an attractive
agenda of labour economists in Pakistan, even though it has been a
common phenomenon in many developing countries. The subject is serious
because it not only helps in understanding workers' behaviour and
decision to allocate his/her time between work and leisure, moonlighting
also affects the very structure of labour market including workers'
performance and productivity. An in-depth analysis is therefore,
required to know the significance of wage rate, demographics of the
labour force, their budget constraints, their engagement in primary and
secondary occupations, and above all their human capital
characteristics. Given this ambitions agenda, it may not be possible to
explore all these issues related with moonlighting in a single study.
Therefore, the present study seeks to focus on three main research
questions. The first and the main objective of the study is to find out
the main determinants of dual job holding; Second, to explore the
demographic and human capital characteristics of moonlighters; and
lastly, to investigate relationship between workers' main
occupation and second job. Within this perspective, the study may be
regarded as a starting point to understand the complexities of labour
market in Pakistan, particularly the job mobility, labour market
transition and spillover over effects, if any.
The organisation of the study is now outlined; next section will
provide a review of background studies followed by a section on the
facts regarding data and its characteristics. The fourth section
presents brief description of the model and methodology adopted in the
paper. The fifth section comprises of empirical results and discussion,
last section concludes the study based on empirical findings of this
paper.
2. BACKGROUND
Despite the importance of dual job holding in today's economy,
it is difficult to find comprehensive information particularly in terms
of empirical evidence in the available economic literature. Despite the
fact that dual job holding in labour market is be the result of many
integrated reasons, Shishko and Rostker (1976), O'Connell (1979)
and Krishnan (1990) have concentrated only on the constraint motives
which restrict the working hours on the primary job and limits the
earning capacity. Every rational worker who wants to maximise his/her
utility would opt for second job if he/she is not satisfied with hours
worked on first job and therefore earning less than his reservation
wage. Paxson and Sicherman (1994) characterise dual-jobs and dual-job
holders, with a focus on dynamics. The aim of their study was to
understand why and when workers moved into and out of second jobs.
Kimmel and Conway (1995) presented a diverse work which examined the
characteristics of moonlighters and also the length of this episode. The
analysis of their article reveals that most moonlighters, in spite of
working long hours, tended to be poorer than the average worker. Berman
and Cuizon (2004) placed multiple job holdings in the context of health
systems and government policies in low and middle-income countries. The
paper offered guidance on how policy-makers could deal with both the
positive and negative view of multiple job holdings. A bivariate probit
model of the decision to work and the decision to hold more than one job
was estimated by Averett (2001), where she found that there was no
difference in factors that influenced the decision to moonlight either
by men or women.
3. DATA
The study exploits the Labour Force Survey 2006-07 to understand
the dynamics of dual job holding in Pakistani Labour Market. To explore
this issue we have restricted this study only to male workers residing
in urban areas. The reasons for these restrictions are based on
different characteristics of rural labour markets; moreover in many
cases females are involved in moonlighting but are under reported thus
not included in estimation. To capture the residential effect the
representation of four provinces is included in the analysis. Since the
main aim of the study is to find determinants of moonlighting and
relationship between primary and secondary occupation. Thus nine
occupational categories along with total hours spent in the labour
market are also included in model specification. The definition of the
variables and their magnitude both for single and dual job holder are
given in Table I.
The total sample comprises of 17248 male workers living in urban
areas, among those 1.3 percent reported as moonlighters. Average age is
high for those with dual jobs holding may be due to increasing
responsibilities at higher age. Similarly, 90 percent of the dual job
holders are married. The average weekly time spent in labour market is
definitely high for moonlighters. The data shows that moonlighters are
endowed with higher number of years of schooling. The Labour Force
Survey provides educational information in the form of different levels
of schooling, which is converted into a continuous variable with number
of years of schooling required for each educational level. The raw data
indicate that the incidence of moonlighting is highest in Punjab and
lowest in Balochistan. The total proportion of workers is high in Sindh
but the practice of moonlighting is higher in Khyber Pakhtunkhwa.
Among the nine occupational categories; managers, technicians and
skilled categories have the highest number of moonlighters. The
proportion of these categories is low as compared to other categories in
overall labour force because these occupation requires specific human
capital endowment. The individuals with the specific human capital
characteristics are low in supply but demand is high in labour market;
thus these people can get involve themselves in multiple jobs in order
to utilise their human capital to its maximum level. Thus the overall
labour force comprises a small proportion of these occupations but
moonlighter are more concentrated in these categories.
4. MODEL AND METHODOLOGY
The moonlighting is supply of labour in more than one job. Even
though there can be different motives for dual job holding, but labour
economists believe (1) that the most important reason to moonlighting is
constraints on working hours in the first job. However the issue becomes
complicated once individuals are engaged in two different types of
occupation. Thus keeping heterogeneous occupations is also the agenda of
this paper.
We start the model using the microeconomic foundation whereby
workers drive utility income (earned and non-earned) and leisure.
Mathematical it can be stated as,
Max u(Y, l) ... ... ... ... ... ... ... (1)
Subject to: [h.sub.1] + [h.sub.2] = 24 - l and [w.sub.l][h.sub.l] +
[w.sub.2][h.sub.2] = Y
Where [h.sub.1], [h.sub.2], [w.sub.1], [w.sub.2], l, and Y are
hours worked and earnings from primary and secondary job (2), leisure,
and total income earned from both jobs. By substituting these
constraints in the utility function we will obtain:
Max u([w.sub.1][h.sub.l] + [w.sub.2][h.sub.2] 24-hl- h2,) ... ...
... ... ... (2)
The constraint of minimum working hour has been applied in
estimation. Thus only those individuals working more than 36 working
hour (3) are included in the estimation. The estimation is based on two
models, first is probit model to examine the determinants of
moonlighting and second model is estimated for the treatment of
endogenous regressor. For the correction of possibility of endogeneity
instrumental variable probit model is used, which adopts two-step
estimation methodology. First model incorporates the effect of total
working hours on moonlighting and second model examines the effect of
weekly wages from the main occupation on moonlighting.
Model 1
The probability of holding two jobs is estimated through following
model specification, where the dependent is a binary variable and thus
probit model is used:
log (Prob(moonlight)/Prob(will hold one job only)) = X[beta] ...
... ... (3)
Or
(Prob(moonlight)/Prob(will hold one job only) = [e.sup.X[beta]] ...
... ... ... (4)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)
The variable determining the probability of holding dual jobs
include age, age square, marital status, number of years of schooling,
four provincial categories and nine occupational categories are included
in this model specification.
Model 2
When any of the regressors (wages in primary occupation in this
case) are endogenous, then estimates are inconsistent [Yatchew and
Griliches (1985)]. For this Instrumental Variable Probit (IVP) model
method is preferred. The parameters are estimated by maximising the
conditional likelihood. This method earlier used by Murphy and Topel
(1985) for such type of correction. The following notation has been
developed to understand the particular model specification used here.
The original model for the decision of moonlighting is as follows:
Pr(Y=1 or Zero) = [beta]T + [gamma] X + [eta]W+ [mu] ... ... ...
... (6)
Where T is age, age square, number of years of schooling, X is for
marital status, provincial dummies and nine occupational variables. W is
for log of weekly wages earned from main occupation. Since wage is
endogenous variable, which it-self is determined by age and schooling,
therefore it may lead towards biased estimates. The problem of
endogeneity is avoided through a two-step instrumental variable probit
model method. In first stage, log of wages are estimated as a function
of age, age square and number of years of schooling.
W = [delta]T + V ... ... ... ... ... ... ... (7)
Where T includes age, age square and schooling, V is an error term.
A new predicted variable is generated (log of weekly wages) based on
estimated co-efficient of T. In the second stage, the predicted values
of log of weekly wages are included in the probit model with rest of
demographic, occupational and regional variables. Finally the model (6)
gets the form presented in Equation (8). In the final Equation T do not
enter in the model rather it enters through predicted wages.
Pr(Y=1 or Zero) = [gamma]X + [eta][W.sub.(predicted) + [mu] ... ...
... ... (8)
The usual probit maximum likelihood procedure is used for the
estimation of the parameters.
5. EMPIRICAL RESULTS AND ANALYSIS
Table 2 presents the results of both models i.e., Probit and two
step IV Probit model. In both models the dependent is a binary variable
and shows the decision of an individual that he/she will moonlight or
not. The estimated co-efficient of the model 1 shows that age and being
married increases the probability of holding two jobs. The probability
of married individual to moonlight is .360 percentage points more as
compared to singles. As far as schooling is concerned it is not only
insignificant but also very small in magnitude, the reason may be that
decision of holding a second job is not necessarily influenced by level
of education. There are many illiterate workers in labour market those
who are engaged in multiple jobs within elementary occupations. Many
workers associated with blue-collar low skilled occupations like
security guards, gardeners, drivers, chefs etc are holding more than one
job. Secondly it is also evident from summary statistics that average
years of schooling is 8.2 years for single job holders and 8.5 for those
moonlighters with a comparatively high standard deviation, thus labour
supply decision of decision of holding more than one job is not
influenced by schooling.
Among all the provincial dummies the probability of moonlighting is
higher in rest of three provinces as compared to Balochistan. The most
plausible reasons are low rate of economic activity in the province and
the proportion of professionals, technicians and related categories
where chances of moonlighting are high comprises very low in total
labour force in Balochistan. Most of the people are concentrated in
elementary occupations of craft related activities. The odd ratio for
Khyber Pakhtunkhwa is highest; reason may be the established and vibrant
business community, huge share of informal sector, trade and related
occupational categories.
We have used occupational categorisation according to international
classification system of occupation. 'Professional' category
is used as base category, the estimated coefficient shows that the
probability of moonlighting is highest for professionals as compared to
any other occupation. Annexure A illustrates the sub-categories of
professional category; which include science and engineering
professionals, health professionals, teaching professionals, business
and administration professionals, information and communications
technology professionals, legal, social and cultural professionals. The
occupational categories require some specific type of education and are
high in demand in labour market. Doctors, engineers, lawyers, and
business professionals are most popular labour market participants
involved in moonlighting. All occupational categories have negative
signs and significant except skill category which include occupations
related with agriculture and fishery industry; in these occupational
categories chances to moonlight are rare.
Another important motive for holding second job is lower wage from
primary job as compared to individuals' reservation wage. The wages
earned during a week is function of age, experience and schooling; due
to the endogenous nature of wage this study used two-step IV Probit
model. Many papers propose to use this mode! in case of endogenous
regressors in probit model [see for example: Arendt and Holm (2006)]. In
second model the significance and signs of all estimated coefficients
remain same; however there are some considerable changes in magnitudes.
The log of weekly wages is a predicted variable and it is function of
age, age square and number of years of schooling. The log of predicted
weekly wages earned from primary occupation are insignificant and are
low in magnitude, which suggest that as the income earned from the
primary occupation do not playa significant role in moonlighting
decision. The result is also confirmed by summary statistics Table 2
that average monthly earnings from primary occupation of single job
holders is almost same of those have dual jobs with a very high standard
deviation. Thus results show that decision to moonlight is not
influenced by income from primary occupation, which shows that there are
some other reasons are more important. The most probable reason is
constraint on working hours on primary job.
The details of major occupational groups along with their
sub-groups are given in Annexure A. Among those moonlighters in
managers' category 46.8 percent are holding their second job in
market-oriented agriculture and fishery related occupation.
Professionals and technicians are only two primary occupations in which
dual job holders are moonlighting within their main occupation.
According to summary statistics presented in Table 3, there is very
small proportion of clerks in the overall labour market. Within those
small numbers of workers in this category those with holding two jobs
mostly are engaged in first occupational category. Those in
'service' category are more related with craft and related
trade occupations in their secondary jobs. Moonlighters both in
'craft' and 'plant' are mostly associated with
'skill' category for their secondary occupation. Thus those
workers who opt for different occupation for moonlight, to investigate
the spillover effect may be an interesting future research agenda.
Now coming back to the main research questions described earlier in
the study. Overall the results show that the being in certain
occupations increases the probability to moonlight. There can be number
of reasons for this, for example under utilisation of skills and
restriction on number of working hours in their primary occupations. The
second main question of the study is to explore the characteristics of
the moonlighters, the reveals that average age of moonlighters is 40
years, 90 percent are married, on average they have nine years of
schooling and 50 percent of total moonlighters are residing in Punjab.
Managers, professionals, technicians and elementary occupations are most
popular among the moonlighters, or in other words we can say that
moonlighters are concentrated on both ends of the occupational
distribution. Lastly, the study finds that managers usually moonlight in
their own occupational category or in skill category, (4) professionals
moonlight in their own primary occupation, technicians also moonlight in
their own occupation but the evidence shows that they also moonlight in
managers and professional category.
6. CONCLUSION
The study has investigated the dynamics of moonlighters and
association between the primary and secondary jobs in Pakistani labour
market while exploiting information from Pakistan Labour Force Survey
2006-07. Surprisingly the wage rate is not the motivation for incidence
of moonlighting, rather individuals reported as moonlighter earning more
as compared to those who are relying on one job only. The last part of
the analysis presents the association of occupation between primary and
secondary occupation, the occupational association analysis shows that
only professionals and technicians are two occupational categories where
moonlighters are holding their secondary jobs in the same occupation.
The study is constrained by many limitations like unavailability of
information on many important variables for example constraint of hours
on primary job etc. But the implications of the study are very
important; it open many venues for further research on this topic, for
example the effect of moonlighting on productivity on first job, labour
market transition, job mobility etc.
Annexure A
INTERNATIONAL STANDARD CLASSIFICATION OF OCCUPATION: MAJOR GROUPS
AND SUB-MAJOR GROUPS
1. Managers
11 Chief executives, senior officials and legislators
12 Administrative and commercial managers
13 Production and specialised services managers
14 Hospitality, retail and other services managers.
2. Professionals
21 Science and engineering professionals
22 Health professionals
23 Teaching professionals
24 Business and administration professionals
25 Information and communications technology professionals
26 Legal, social and cultural professionals.
3. Technicians and Associate Professionals
31 Science and engineering associate professionals
32 Health associate professionals
33 Business and administration associate professionals
34 Legal, social, cultural and related associate professionals
35 Information and communications technicians.
4. Clerical Support Workers
41 General and keyboard clerks
42 Customer services clerks
43 Numerical and material recording clerks
44 Other clerical support workers.
5. Service and Sales Workers
51 Personal service workers
52 Sales workers
53 Personal care workers
54 Protective services workers.
6. Skilled Agricultural, Forestry and Fishery Workers
61 Market-oriented skilled agricultural workers
62 Market-oriented skilled forestry, fishing and hunting workers
63 Subsistence farmers, fishers, hunters and gatherers.
7. Craft and Related Trades Workers
71 Building and related trades workers, excluding electricians
72 Metal, machinery and related trades workers
73 Handicraft and printing workers
74 Electrical and electronic trades workers
75 Food processing, wood working, garment and other craft and
related trades
workers.
8. Plant and Machine Operators, and Assemblers
81 Stationary plant and machine operators
82 Assemblers
83 Drivers and mobile plant operators.
9. Elementary Occupations
91 Cleaners and helpers
92 Agricultural, forestry and fishery labourers
93 Labourers in mining, construction, manufacturing and transport
94 Food preparation assistants
95 Street and related sales and service workers
96 Refuse workers and other elementary workers.
REFERENCES
Averett, Susan (2001) Moonlighting: Multiple Motives and Gender
Differences. Applied Economics 33:11, 1391-141.
Arendt, J. N and A. Holm (2006) Probit Models with Binary
Endogenous Regressors. University of Copenhagen. Department of
Economics. Centre for Applied Microeconometrics in its series CAM
Working Papers with number 2006-06
Berman, P. and D. Cuizon (2004) Multiple Public-Private Jobholding
of Health Care Providers in Developing Countries: An Exploration of
Theory and Evidence. Department for International Development Health
Systems Resource Centre.
Krishnan, Pramila (1990) The Economics of Moonlighting: A Double
Self-selection Model. Review of Economics and Statistics 72:2, 361-67.
Kimmel, J. and Powell (1999) Moonlighting Trends and Related Policy
Issues in Canada and the United States. Canadian Public Policy 25:2,
207-31.
Kimmel, J and K. S Conway (1995) Who Moonlights and Why? Evidence
from the SIPP, Upjohn Institute. (Staff Working Paper 95-40).
Murphy, Kevin M. and Robert H. Topel (1985) Estimation and
Inference in Two-step Econometric Models. Journal of Business and
Economic Statistics 3:4, 370-379.
O'Connell, John F. (1979) Multiple Job Holding and Marginal
Tax Rates. National Tax Journal 32:1, 73-76.
Paxson, Christina H. and Nachum Sicherman (1994) The Dynamics of
Job Mobility and Dual-job Holding. NBER (Working Paper No. 4968).
Shishki, Robert and Bernard Rostker (1976) The Economics of
Multiple Job Holding. American Economic Review 66:3, 298-308.
Stinson, J. F. (1990) Multiple Jobholding up Sharply in the
1980's. Monthly Labour Review 3-10.
Yatchew, Adonis and Zvi Griliches (1985) Specification Error in
Probit Models. 'The Review of Economics and Statistics 67: 1,
134-139.
(1) A verett (2001), Kimmel and Powell (1999), Krishnan (1990),
Paxson and Sicherman (1994), Shishki and Rostker (1976) and Stinson
(1990).
(2) "Y" refers only earned income, non-earned income is
not included in out model specification.
(3) According to the definition of ILO those working less than 36
hours are considered as part-time workers. Since the purpose of paper is
to explore the issue of moonlighting thus we excluded those working less
than 36 hours. Reference:
http://www.ilo.org/public/english/protection/condtrav/pdf/infosheets/
wt_4.pdf
(4) Skill category mostly comprises of agriculture related
occupations.
Asma Hyder <
[email protected]> is Assistant
Professor and Ather Maqsood Ahmed <
[email protected]> is
Professor in Economics Department at the NUST Business School,
Islamabad.
Table 1
Definition and Mean of the Variables
Variables Definition Single Job Moon-
Holders lighters
Age Age in complete years 34.52 39.94
(11.58) (10.301)
Hrwork Hours spent during one week in 54.15 60.73
Labour Market (Primary + (10.67) (11.31)
Secondary)
(S.D)
Wkearn Monthly income earned from 1383.99 1425
primary occupation (1562.28) (1417.26)
(S.D)
MS Marital Status .6837 .9056
School Number of years of schooling 8.2373 8.56
(S.D) (5.4053) (6.02)
Punjab Dummy if residence is in Punjab .466 .4979
Sindh Dummy if residence is in Sindh .3075 .1784
Khyber Dummy if residence is in Khyber .1147 .2940
Pakhtunkhwa Pakhtunkhwa
Balochistan Dummy if residence is in .11014 .0290
Balochistan
Manager Dummy if Occupational category .2736 .1950
is Manager
Professional Dummy if Occupational category .0326 .1161
is Professional
Technical Dummy if Occupational category .0619 .1410
is Technical Worker
Clerks Dummy if Occupational category .0371 .0622
is Clerical Staff
Service Dummy if Occupational Category .1007 .0705
is Service
Skill Dummy if Occupational category .0417 .1244
is Skilled Worker
Craft Dummy if Occupational Category .2281 .0788
is Craftsman
Plant Dummy if Occupational Category .0730 .0663
is Plant Operator
Elementary Dummy if Occupational Category .1509 .1452
is Elementary
Total Sample 17248 241
Table 2
Probit Result (Dependent Variable: Moonlighting = 1 if
Moonlight and Zero Otherwise)
Model 1 (Probit Model) Model 2 (IV Probit Model)
Co-efficient Co-efficient
Variables (Std.Errors) (Std.Errors)
Age .0361 * --
(.020)
Age Sq -.00037 --
(.0002)
MS .3608 *** .5645 ***
(.1021) (.0849)
School .0059 --
(.0057)
Punjab .6196 *** .5635 ***
(.1372) (.1291)
Sindh .3762 ** .3138 *
(.1443) (.1366)
Khyber Pakhtunkhwa .9621 *** .9176 ***
(.1430) (..1347)
Manager -.7053 *** -.6837 ***
(.1128) (.1116)
Technical -.1969 * -.1741
(.1209) (.1201)
Clerks -.3454 * -.3503
(.1452) (.1451)
Service -.6232 *** -.6047 ***
(.1402) (.1383)
Skill -.1282 -.0422
(.1384) (.1338)
Craft -.9214 *** -.8722 ***
(.1334) (.1306)
Plant -.6216 *** -.5895 ***
(.1463) (.1440)
Elementary -.5709 *** -.5325 ***
(.1303) (.1271)
Lnwage (predicted) -- .041
(.0810)
Constant -3.25 *** -2.967 ***
(.4024) (.6187)
No. of Observations = 17489 No. of Observations = 17489
Wald Chi2(15) = 263.74 Wald Chi2(15) = 391.68
Prob > Chi2 = .0000 Prob > Chi2 =.0000
Log Likelihood =-1140.03477 Pseudo R2 = .1097
Pseudo R2 = .1037 Log Likelihood =-1170.841
Table 3
Occupational Association between Primary Occupation and Secondary
Occupation
Sec-Occupation Mana- Profes- Tech-
Main-Occupation ger_2 sional 2 nical_2 Clerk_2 Service_2
Manager_1 12.7% 3.12% 10.63% 0% 3.12%
Professional_1 7.14% 67.85% 21.42% 0% 0%
Technical_1 14.28% 14.28 37.14% 0% 0%
Clerks_1 53.33% 0% 6.66% 0% 6.66%
Service_1 11.76% 5.88% 5.88% 0% 5.88%
skill_1 26.47% 0% 5.88% 0% 0%
Craft_1 20% 0% 5% 0% 0%
Plant_1 31.25% 12.5% 6.25% 0% 0%
Elementary_1 11.42% 0% 2.85% 0% 5.71%
Sec-Occupation Elemen-
Main-Occupation Skill_2 Craft_2 Plant_2 tery_2
Manager_1 48.8% 17.02% 0% 4.2%
Professional_1 0% 3.57% 0% 0%
Technical_1 11.42% 11.42% 5.71% 5.71%
Clerks_1 13.33% 6.66% 13.33% 0%
Service_1 11.76% 29.41% 5.88% 23.52%
skill_1 17.64% 0% 0% 50%
Craft_1 45% 25% 5% 0%
Plant_1 37.5% 12.5% 0% 0%
Elementary_1 31.42% 0% 2.85%r 45.71%
