Manpower planning in Pakistan: statistical pitfalls.
Herman, B. ; Irfan, Mohammad
INTRODUCTION
There have been manpower planning exercises in Pakistan with
varying levels of quality and rigor. These exercises are reviewed in
Herman and Irfan (1989) and Kemal (1987). For a comparison with other
countries, one can also refer to Amjad (1987) wherein these exercises
carried on in various Asian developing countries are reviewed. Most of
the exercises suffer from predictive errors as revealed by post-fact
comparison and the studies in this field generally lament about the
non-availability of adequate data, which could be considered as a major
limitation in this respect.
In this paper, an attempt is made firstly to extend the past
simulations such as to correspond with the time horizon of the
Perspective Plan, i.e., by the year 2003. Secondly, in order to
demonstrate the usefulness of these exercises, some policy simulations
are made and their effect on unemployment is assessed. Finally, rather
than warning the reader about the inadequacy of the data, this paper
tries and identifies the main areas wherein statistical efforts must be
concentrated upon for improvement. In order to indicate clearly to the
reader where, in the authors' opinion, are the statistical
shortcomings of a simulation exercise such as the one described in this
paper, it is deemed better to present first of all the general framework
supporting the simulation. On the basis of this scaffolding, the
location of the pitfalls will manifest themselves.
The structure of this type of model is most simple and it can be
graphically visualized through Fig. 1.
THE DEMAND SIDE
The Demand side recognizes two sorts of labour demands, namely,
labour demanded for productive activities such as agriculture,
manufacturing, etc. (to be called workers hitherto), and labour demanded
by the educational sector, (to be called teachers hitherto).
[FIGURE 1 OMITTED]
Labour requirements, in both instances, are considered to be
technically given; in case of productive labour, the number of workers
required by the various sectors for the various occupational classes are
determined by sectoral value added while the number of teachers required
by the various levels of the educational system are determined by the
enrollments at each level.
For simulation purposes, in case of workers, the value added of the
various sectors is supposed to be growing according to the sectoral
growth rates indicated in the Seventh Five-Year Plan; in case of
teachers, enrollments can be variously supposed to be growing according
either to demographic growth or to growth planned by the Ministry of
Education or alternative plans.
THE SUPPLY SIDE
In contradistinction with the Demand side, the Supply side is
supposed to be behaviouristically instead of technically determined. The
modelling of the supply side consists of three sub-models, namely, a
demographic sub-model (indicated in the lower box of the above Graph as
Demo), an educational sub-model (indicated in the lower box as
Education) and a labour market sub-model (indicated in the lower box as
MP).
The demographic sub-modal calculates per cohort the number of
children delivered by society to the educational system. Societal
behaviour is reflected in the various demographic rates.
The educational sub-model is an algorithm describing the
educational pipeline. It is based on a parametric infrastructure
consisting of the following parameters: enrollment rates, internal
efficiency of the system rates (namely, drop-out, repetition and
graduation rates) and the forward rates. On the basis of all these
various rates, the transition and the retention rates are obtained
ex-post instead of being imposed ex-ante, procedure which permits more
sensitive and more varied simulations.
The labour market sub-model calculates the gross potential new
entrants into the labour force and on the basis of parametric
descriptions of the labour market mechanism (i.e., participation rates,
etc.), it delivers the simulated labour supplies by level of education.
These labour availabilities are then to be compared with the
requirements described above.
PROJECTIONS OF LABOUR AVAILABILITIES AND REQUIREMENTS
Concerning population growth, recent information provided by the
last Economic Survey indicate that actual rates are higher than 3.15
percent per annum and growing. For the purpose of these simulations,
population was supposed to grow at 3.15 percent per annum in the DEMO
version and at 3.35 percent per annum in the UPE version.
The basic run assumes that while the parametric infrastructure of
the model remains unchanged, the educational system increases only
because of demographic growth (henceforth referred to as DEMO version).
The above sub-model delivers the gross potential new entrants into
the labour force. Information on participation rates per age cohort and
educational attainments are simply not available. Recent information
published in the last issue of the Economic Survey points out towards
the fact of decreasing participation rates of those age cohorts of
school-going youth. After discussions and consultations it was decided
to adhere to purely fictitious rates, namely, 0.27 for younger and 0.47
for older boys. Table 1 provides the labour availability for different
years by broad educational groups, under DEMO version assumptions.
LABOUR REQUIREMENTS
In a nutshell, sectoral labour demands per occupational class are
estimated as a function of sectoral value added. For matters related to
regression analysis, estimation, significance and plan simulations by
means of the RAS method, the interested reader is referred to Herman
(1989), where labour demands per sector and occupation as well as
sectoral labour demands and national labour demands per occupation are
given.
Recent information delivered by the last issue of the Economic
Survey indicate achievements in labour productivity. This means that
labour absorption is declining. Labour absorption is easily captured by
the income elasticities of employment; indeed, these indicate the growth
of employment which corresponds to one percent growth of income.
Elasticities, which secularly remained above 0.4, are now around 0.35.
In order to incorporate this new information, the matter arises whether
these are to be interpreted as cyclical or structural changes and, in
case of the latter, whether they are to be interpreted as short or as
long-run structural changes.
An eclectic approach is followed in the simulation reported here
below, by which, resulting elasticities will steadily decrease from 0.49
in 1988-89 through 0.4 around 1990 and 0.38 in 1991-92, to stabilize on
0.34 from the end of the Seventh Five-Year Plan period onwards.
Labour demands for productive activities per educational
attainments aggregated over occupations are shown below, in the
Appendix, as stocks as well as in terms of annual increments.
Concerning demand for teachers, it was mentioned already that it is
determined as a function of enrollments. The functional relationship is
most simple; indeed, the number of teachers is obtained multiplying the
number of those enrolled by the student/teacher ratio. For the Primary
level, this ratio is taken to be equal to 35 and for Secondary and
higher levels, to be equal to 30. Although these values might seem to be
too low, the rationale was that for improvements of the educational
system as well as for increasing the absorption of educated youth, a
reduction of the student/teacher ratio was deemed to be desirable.
The resulting numbers of teachers required per educational level
are available on request. Adding over educational levels, one obtains
that by 1989 about 320 thousand teachers are required and that by the
end of the plan period these numbers, in the DEMO version, will have to
increase by 50 thousand more, that is, roughly speaking, an annual
increment of 10 thousand teachers per year. The UPE version, of course,
shows a completely different picture since enrollments, besides
demographic growth, are simulated to increase because of policy. Indeed,
were all 5-year-olds enrolled in first class Primary school by 1995, the
number of teachers will have to increase from 320 thousand at the
beginning of the Seventh Five-Year Plan to 460 thousand at the end of
the plan period and up to 660 thousand four years later which implies
annual increments of about 45 thousand extra teachers per year.
IMBALANCES
In order to reflect the imbalances, a measure was built equal to
the difference between supply and demand expressed as percentage of
supply. This is roughly speaking equivalent to the so-called
unemployment rate. The imbalances yielded by the basic DEMO run are
provided in Table 2. (Teachers demand included in total demand; further,
usual DEMO version assumptions).
It needs to be mentioned that the imbalances have been arrived at
by incorporating in the demand side the requirements for teachers
assuming a teacher/ student ratio which does not prevail at the moment.
In other words, these imbalances reflect already a policy measure. It
may, however, be observed that the overall unemployment level rises from
5 percent in 1989 to 14 percent in the year 2003. In terms of flows, it
means that 40 percent of the new entrants into the labour force will not
be absorbed. A slightly lower level of unemployment among the educated
than the overall employment rate reflects the effect of the assumptions
made regarding the teacher/student ratio. In case one ignores the impact
of demand for teachers and considers only the productive system's
demand for educated labour, the unemployment rate for the educated jumps
from 13.4 to 22.1 percent while 47 percent of the new entrants will be
unemployed by 2003, instead of 37.6 percent.
POLICY SIMULATIONS
Another interesting policy simulation to highlight the divergence between the growth pattern and rising unemployment among educated youth
was made wherein the system was forced to absorb the educated unemployed
to the extent that for this group, the 4 percent unemployment rate is
not exceeded in 1992-93. Interestingly, the required growth rate of the
economy would be 15 percent per annum rather than 6 percent as planned
for the Seventh Five-Year Plan. Unfortunately, neither the resulting
combination of workers nor the implied sectoral growth rates can be
obtained as yet unless the model is made to be reversible.
Effect of Universalizing Primary Education, on Unemployment
In this paper we are presenting results of still another policy
simulation which reflects the achievements of an hypothetical
universalization of primary education. Under this policy, there is an
increase in enrollments as well as in the teacher requirements. It might
be remembered that in the basic rum, while the parametric infrastructure
of the model remains unchanged, the system increases only due to
demographic growth (referred to as DEMO version). Under universalization
of primary education (UPE version), one of the many possible simulations
was tried out, namely, what the authors want to call the Baqai
simulation in honor of Dr Moinuddin Baqai. The policy implied lets
enrollment of 5-year-olds to steadily increase from 53 percent in 1989
to 100 percent in 1995. All other rates and parameters are left
untouched.
The resulting UPE version shows a different picture. While labour
demands of both types are shown in Appendix Table 3, here below in Table
3 the implied imbalances are presented.
A comparison of Table 3 with Table 2 suggests that unemployment
both amongst illiterates as well as educated, is significantly reduced
(from 13.4 to 6 percent for the educated). However, increasing the
number of teachers only for absorption of educated labour, even in the
extreme case of UPE, is a remedy that exhausts itself in a decade or so
(educated unemployment decreases up to a minimum around 1995 and then
increases again). This underscores the need for recognizing these
measures as mere palliatives which simply provide a short-term breathing
space. Still, one cannot underrate these measures, if not to solve
educated unemployment, indeed to improve the educational system, even at
the risk of generating more educated youth. Actually, only by engaging
more educated workers in the production process, can unemployment be
tackled significantly. It is a curious fact that although educated
unemployment seems to be a problem bound to stay, the absolute numbers
of educated youth generated by the educational system are extremely low
(150 thousand in a country of 110 million!).
STATISTICAL PITFALLS
The foregoing discussion amply demonstrates that manpower planning
is a useful and policy relevant exercise. In order to enhance its
reliability and minimize predictive errors, one needs to concentrate
efforts to make available a more reliable set of data than that now at
our disposal.
Without aiming at an exhaustive treatment, for the purposes of this
paper four areas urgently requiring improvement are to be singled out.
These are, one on the supply side of the model, one on the demand side,
another affecting both the demand as well as the supply sides and,
finally, still one more concerning the imbalances block. (See Fig. 1).
PARAMETRIC BASIS OF SPLIT PIPELINE
The educational sub-model located on the supply side consists
essentially of a pipeline receiving children on the one hand and
delivering potential entrants into the labour force, on the other.
Children are enrolled, they may drop-out from, or repeat a class, they
may pass to the next class or graduate from a certain educational level,
then they may join the labour force or go forward to the next higher
educational level, etc.
All these changes in their status are simulated by means of the
corresponding rates. For the purposes of the illustrations shown above,
educated guesstimates were made, based on partial sources of information
on the various rates as well as on statistical sources providing
information on absolute numbers. The guesstimation procedure boiled down
to reconciliate the one with the other, namely, making rates as reported
deliver absolute values as indicated in statistics, and vice versa, that
is, forcing absolute values to abide by rates. The resulting algorithm
is a model description of reality, that is, it abstracts only the
relevant features in order to simulate results under various assumptions
such as policy changes, etc.
This modelling of Pakistan's educational system depicts,
indeed, the situation prevailing in the whole nation. All parameters and
values used are, thus, either national averages and/or national totals.
The usefulness of this modelling exercise is two-fold: first it
generates information providing the national context and, second, it
provides the skeleton of an educational system, whichever it may be.
It must be well understood that national averages subsume wildly
diverse situations at the disaggregated level. Indeed, average drop-out
rates or average enrollment rates (or, in general, each of the rates
built-in in the algorithm) do not reflect the circumstances and the
differences existing between, say, the urban and the provincial
sub-systems, or between boys and girls.
It is thus imperative to conduct research addressed towards
disaggregating the national averages showing at least the above
mentioned categories. In this way, instead of one single pipeline, one
would be able to build several. Indeed, four pipelines would be the bare
minimum, i.e., one for boys and one for girls, as well as urban and
rural ones. Only thinking in terms of solving imbalances, these split
pipelines will already manifest their usefulness because, e.g., excesses
of male teachers in urban areas cannot be used to compensate shortages
of female teachers in rural areas, etc.
THE LABOUR DEMAND EQUATION
Although, for the purposes of a manpower planning exercise, it is
well recognized that requirements of any one factor (in this case,
Labour) are to be estimated by means of a simultaneous model solving for
requirements of all factors, for their remunerations and for the level
of production, the condition imposed to disaggregate made this orthodox
venue to be impossible. Such an approach would have meant the solution
of a huge general equilibrium model whose solution is consistent with
the solution of nine sectoral equilibrium models each of which
consisting of fifty production factors (i.e., forty-nine sorts of
labour--seven times seven-plus capital) and fifty factor prices. Besides
the theoretical difficulties, this venue was dismissed simply because of
the scarcity of data.
Still, keeping in mind the structure of the above general
equilibrium formulation, it is possible to think in terms of a reduced
form equation for labour utilization where levels of production, of
utilization of other factors, of remunerations to each factor as well as
lagged levels of utilization of all factors are exogenously given. This
reduced form equation for labour demand will come specified according to
the specifications adopted for the various equations of the original
general equilibrium model and will be expressed in terms of combinations
of parameters such as the distribution coefficient, the substitution
elasticity, the rate and class of technological growth, the
(price)-income valuation of output, depreciation and other factor
replenishment coefficients, etc.
A simplification of such a reduced form equation for labour
utilization, is the conventional labour demand equation. This is a
technical equation determined empirically. It is technical since it is
based on the assumption that there is a technical necessity of labour to
attain a certain income, and it is empirical since its parametric
specification, subsuming all the above mentioned parameters, is chosen
such as to fit the data.
It must be well understood, however, that the above described
simplification is no more than that, a simplification. It, in fact,
perverts the results by subsuming all the undercurrents that motivated a
certain situation which was, afterwards, quantified in the statistics.
The researcher actually observes the surface without entering in the
mechanisms that delivered such a surface. These mechanisms are those of
the labour market and are to be investigated.
A compromise might provide a way out. Thinking in terms of a
general equilibrium model solved at macro and at micro level, the
researcher might attempt to disentangle the working of the labour market
mechanism. Results will anyway be macro, no information will be provided
on occupations, the micro level will remain obscure, etc. But,
nevertheless, taking into consideration that manpower planning is a
macro exercise, such a compromise solution will relate manpower demands
to levels of output, taking into consideration demands for other factors
of production, their substitution possibilities, as well as the factor
prices of all factors involved. The shortcomings of the conventional
labour demand equation will be removed to a large extent.
EDUCATIONAL PROFILES
This is a typical statistical pitfall. It affects both sides of the
problem, namely, the demand as well as the supply side. Indeed, once
that labour requirements were obtained per sector of production
according to expectations concerning sectoral output levels, these
labour demands must be expressed in terms of demands per occupational
classes in order to enter into the manpower planner's proper field
of action. This conversion can be done since some information is
available describing the occupational structure of employment per
sector. But entering into his field of action is not enough: the
manpower planner must dwell in it and derive conclusions. To do that he
must disentangle the mechanisms which delivered workers endowed with the
necessary skills as to discharge their tasks in their respective
occupations they are holding. For that purpose, the educational profiles
of the occupations are needed.
Statistical work aimed at mending this situation must begin by
improving the currently used occupational classification. Indeed, a
classification which permits large number of workers to be classified in
one or another class just by drafting them into the bunch of
"Related activities" is bound to deliver confusing results.
INVESTMENTS TRADE-OFF IN PRODUCTION VERSUS EDUCATION
Labour demands determined by expected expansions in production
which would result from planned investment allocations may not match
labour supplies. These labour supplies, per segment of the labour
market, are the results of decisions pertaining investment allocations
embodied in the labour force. Given that the total available fund for
investments is predetermined, a way towards clearing the segmented
labour markets would require making the system to be reversible. That
is, instead of asking what would the imbalances be which would result
from investment decisions allocating funds to both production and
education, one would rather need to ask which need the investment
allocations be that would result in the smallest imbalances.
In Herman and Irfan (1989) a proposal was put forward which permits
running the model back and forth, as many times as it is necessary to
clear each of the segmented markets without tampering with the rate of
growth of the national economy.
As already said, this might be a way out. Running manpower planning
exercises omitting this approach may generate results falling far off
the mark because labour markets do work, with or without the acquiesence
of the manpower planner. The problem is that they clear themselves
solving the imbalances in ways that are sub-optimal from the national
point of view and which create unnecessary tensions in the labour force
by engaging in production sub-optimally qualified labourers (either
over-qualified or under-qualified).
Comments on "Manpower Planning in Pakistan: Statistical
Pitfalls"
The paper under review is an interesting manpower planning exercise
with three stated objectives, i.e., to present projections of employment
and unemployment trends to cover the period of the perspective plan to
the year 2003; to present policy simulations to demonstrate the
usefulness of such exercises and finally to identify and highlight the
areas of major deficiencies in data. The paper can be easily divided
into two distinct parts; an excellent critical appraisal of the
statistical pitfalls and the needed improvements, that forms the latter
part of the paper and a rather general application of an extremely
simplistic and highly aggregated manpower planning model and some
simulations that form the first part. The first part raises a number of
questions and issues in the readers mind as to the raison d'etre for the whole exercise and for the need to make specific assumptions.
Some of these questions are answered in the second part. It might,
therefore, have been worthwhile to restructure the paper so that the
second part came first.
The topic is extremely important to Pakistan and the analysis could
have serious policy implications. However, in their efforts to address a
problem of extraordinary complexity with extremely inadequate data and
in a few short pages, the authors have left themselves vulnerable to a
number of noticeable pitfalls. I list below the more obvious ones with a
view to provoking further discussion and perhaps, more detailed analysis
in the future.
The model on the demand side sees the total requirements of
manpower as the sum of the labour demanded for productive activities
such as agriculture and manufacturing etc., and the labour demanded by
the educational sector (teachers). This is an interesting breakdown
because the education sector is generally considered part of the larger
services sector which seems to be missing in the model. Labour
requirements are obtained by aggregating over various sectors the
numbers obtained by applying sectoral labour: value-added ratios. For
purposes of simulation the study assumes that the various sectors are
growing according to the sectoral growth rates indicated in the Seventh
Five-Year Plan. An obvious question that comes to mind is whether these
growth rates assumed in the Seventh Plan were actually met? Would it not
have been more realistic to use actual growth rates rather than those
assumed in the five-year plan.
Labour requirements by educational attainments are obtained in the
modal by imposing educational profiles for various occupations. This
again raises the problem that runs through the entire modelling and
estimation presented in the paper. It is one of imposing static, highly
aggregated and questionable parameters to predict a process that is
dynamic and disaggregated.
The supply side of the model is supposed to be behaviouristically
determined and consists of three sub-models i.e., the demographic, the
educational and the labour market. Each one of these models is extremely
sensitive to the different rates used to depict either societal
behaviour of the demographic model, the parametric infrastructure of the
education pipeline or the parametric descriptions of the labour market
mechanism. Here again the validity of the entire exercise depends upon
the starting values.
The questionable nature of most of the rates used in the
simulations is highlighted by the authors own discussion of the
"purely assumed" decreasing labour force participation rates
of school-going youth namely 0.27 for younger and 0.47 for older boys.
The level of employment, and by default, that of unemployment would
depend crucially on the values assumed for the income elasticity of
employment. The authors having categorically established that the income
elasticity of employment is "now around 0.35" (no source is
cited for this information) go on to use an "eclectic
approach" in the simulations wherein they allow the elasticities to
decline "from 0.49 in 1988-89 through 0.4 in 1990 and 0.38 in
1991-92 to stabilize around 0.34 from the end of the Seventh Plan period
onwards". While the authors are at liberty to assume any values
they like, the implications for total employment (and unemployment) of
an elasticity of 0.49 instead of 0.35 in the starting period are,
gigantic. There are also several errors of "adding up" or
discrepancies between the numbers in the tables and the descriptions in
the text (rounding errors!) that can have implications that when
translated mean either thousands more employed than the study is meaning
to show or thousands more unemployed than the exercise has revealed.
However, it is to the authors credit that they are aware of all
these shortcomings and have hinted at most of them either in the second
half of the paper or in their presentation. It is also to the authors
credit that while qualifying that all parameters are either national
averages and/or national totals they state clearly that it must be well
understood that national averages subsume wildly diverse situations at
disaggregated level. The study makes an earnest call for research
addressed towards disaggregating the national averages to reflect the
differences between urban and rural, male and female and provincial
sub-systems. Till such time as data are available that permit this to be
effectively done, the study under review adds some more evidence, most
of it generated by the two authors of the present study, to the need for
effective manpower planning in Pakistan.
Sohail J. Malik
Pakistan Institute of Development Economics, Islamabad.
Appendix Table 1
Workers Requirements for the Productive System, per Educational
Attainments
Stocks Flows
(Millions) (Thousands)
Year L <M M&up L <M M&up
1988-89 29.9 26.6 3.31 910 809 101
1992-93 32.9 29.2 3.64 693 616 77
1997-98 36.4 32.4 4.03 727 647 81
2002-03 40.3 35.6 4.46 805 716 89
L = Total demand for productive workers, of which.
<M = Workers with less than Matric.
M&up = Idem with Matric and above.
Growth rates of value added planned for the Seventh Five-Years Plan,
are supposed to remain valid beyond the plan period.
Appendix Table 2
Workers and Teachers Requirements per Educational
Attainments; DEMO Version
Stocks Flows
(Millions) (Thousands)
Year L < M M&up L < M M&up
1988-89 30.2 26.6 3.63 920 809 111
1992-93 33.3 29.2 4.01 705 616 89
1997-98 36.8 32.4 4.46 742 647 95
2002-03 40.8 35.8 4.97 822 716 106
Appendix Table 3
Workers and Teachers Requirements per Educational Attainments;
UPE Version
Stocks Flows
(Millions) (Thousands)
Year
L < M M&up L < M M&up
1988-89 30.2 26.6 3.63 920 809 111
1992-93 33.3 29.2 4.10 738 616 122
1997-98 37.1 32.4 4.74 773 647 126
2002-03 41.2 35.8 5.39 849 716 134
REFERENCES
Amjad, Rashid (1987) Human Resource Development. In Rashid Amjad
(ed)Human Resource Planning. Delhi: ILO, Asian Employment Programme.
Herman, B. (1989) Estimation of Labour Demands: A Check and a
Proposal. HRD Note # 1, Islamabad: Pakistan Manpower Institute.
Herman, B., and M. Irfan (1989) Some Employment Simulations for the
Seventh Five-Year Plan. In S. N. Hyder (ed): Human Resource Development
and Utilization. Islamabad: Pakistan Manpower Institute.
Kemal, A. R. (1987) Pakistan Experience in Manpower Planning. In
Rashid Amjad (ed) Human Resource Planning. Delhi: ILO, Asian Employment
Programme.
B. HERMAN and MOHAMMAD IRFAN *
* The first author is associated with the Pak/Dutch HRD Project and
the second author is Chief of Research, Pakistan Institute of
Development Economics, Islamabad.
Table 1
Labour Supply per Educational Attainments (DEMO)
Stocks Flows
(Millions) (Thousands)
Year L < M M&up L < M M&up
1988-89 31.9 28.3 3.58 902 757 145
1992-93 35.8 31.6 4.16 1003 857 146
1997-98 41.2 36.3 4.92 1159 1002 156
2002-03 47.5 41.8 5.73 1338 1170 167
L = Total supply of workers, of which.
<M = Workers with less than Matric.
M&up = Idem with Matric and above.
Table 2
Imbalances = (Supply-Demand)/Supply; (%). (DEMO)
Stocks Flows
Year L M&up L M&up
1989 5.2 -1.5 -2.0 23.9
1993 7.0 3.5 29.8 39.5
1998 10.7 9.3 36.1 40.1
2003 14.1 13.4 38.7 37.6
Table 3
Imbalances = (Supply-Demand)/Supply;
(%). UPE
Stocks Flows
Year L M&up L M&up
1989 5.2 -1.5 -1.7 23.7
1993 6.8 1.4 27.2 16.2
1998 9.9 3.5 29.6 19.1
2003 12.4 6.0 32.6 22.5
