evidence on negative employment e ffects of minimum wages
TRANSCRIPT
Evidence on Negative Employment Effects of Minimum Wages in Inflexible
Markets:
Part-time Jobs and Female Labor Force Participation in Turkey
Incomplete and Preliminary
Orgul Demet Ozturk∗†
Last Updated
Nov 21st, 2003
Abstract
Despite the significant improvements in the social and economic position of women, the fe-
male labor force participation rate has been decreasing in Turkey over the last 50 years. Thispaper attempts to explain this continuous decline in the participation rate by pointing out thenegative employment effects of minimum wage regulations in inflexible markets. When there arefixed costs associated with hiring, the minimum wage destroys part-time jobs for women at thelow end of the productivity distribution since it makes it prohibitively expensive to offer shortworkdays to low productivity workers. Thus, employers include minimum hours requirementsin the contracts they offer. More than half of the non-participant women in Turkey are “house-wives”. The results indicate that married women are the first to be affected by the constraintson the hours offered as they have higher-valued outside options (household production), andprefer to supply lower hours to the market. This paper also offers an explanation for observed
“wage premium” in the existing part-time job market.
Keywords: Female labor force participation, minimum wage, part-time jobs, mini-
mum hours requirement
∗University of Wisconsin-Madison, Economics Department, 1180 Observatory Dr., Madison, WI 53706-1393. e-
mail: [email protected]†I would like to thank John Kennan and Jim Walker for their guidance. I also would like to thank John
M.Gordanier, Seokjin Woo and Sudip Gupta for their support and valuable ideas; Cem Baslevent for helping me with
the data; A. Filiztekin, Lucas W. Davis, Maurizio Mazzocco, Oya P. Ardic, Sibel Sirakaya and William E. Riordan
for their helpful comments. All mistakes are mine.
1
1 Introduction
Contrary to the expectations arising from the demographic and social changes over the last 50 years,
the female labor force participation rate has decreased significantly in Turkey over this period. The
female labor force participation rate was 72 percent in 1955 and it declined to 27.5 percent in 2001.
In 2001, the participation rate was only 16.9 percent among urban women (SIS, 2001).
The initial drop in the participation rate is attributed to the massive urbanization of the work-
force after the 1950s. Small scale, family-level agriculture had been employing nearly all of the
women in rural areas. Since the distinction between household duties and work is blurred in
agriculture, it is easier for rural women to meet the conditions to be considered as employed. It
is argued that when these women leave their villages and move to the cities, they cannot find a
place for themselves in the labor force of urban Turkey (Ecevit, 1990; Dayioglu, 1998; Ozar, 1996;
Tunali, 1997). In cities, market work and household duties are incompatible. Hence, women have
to concentrate on one of them. Most of these women have little human capital, so they are forced
into “marginal” jobs. Faced with this, most choose not to participate in the workforce.
However, the continuing decline in the participation rate is unexpected since the social status
of women has improved significantly over these years. Through vast public programs, increased
emphasis on compulsory schooling, and the introduction of secularity in aspects of social life, the
education level of women has increased substantially. This is accompanied by a considerable drop
in total fertility rates (from 6.6 births per women in the 1950s to 3.3 in 1988, and 2.16 in 2001),
and a gradual increase in the marriage age and the age of the first birth (Shorter, 1995). Previous
empirical studies on experiences of other countries imply that these changes should lead to a higher
female labor force participation rate (Mincer, 1985; Schultz, 1990; Goldin 1995).
There is no study yet that provides a convincing explanation why the Turkish economy is [ever
more] incapable of utilizing the increasing productivity of women. Baslevent (2001) studies the
full-time vs. part-time work decision of women. He proposes that if there were more part-time
jobs, participation in the workforce would be higher. However, this conclusion is not satisfactory.
Why is there a scarcity of part time jobs? How large an impact does scarcity of part-time jobs
have on the participation rate?
This paper aims to fill this gap and to explain the “housewife-ization”(Ecevit, 1990) of urban
females by pointing out the negative employment effects of minimum wage regulations in Turkey1.
1Minimum wage is an old concept in Republican Turkey. It was first proposed, accepted and placed in The Labor
Law in 1936 (Law number 3008-32). However, it was not enforced until 1951 since the required legal regulations had
not been completed in the previous 15 years. In this law, minimum wage was defined as the “wage that is enough
to provide a worker with his basic needs which will be paid to the equal work performed under normal conditions”.
Between 1951 and 1954 minimum wage was determined by local “minimum wage commissions”. Enforcement was
not very wide though: minimum wages were regulated only in 12 cities out of 67.
2
Restricting the participation analysis only to the supply side is the reason most of the previous
studies failed to adequately explain the big picture. This paper models the demand side to explore
the workings of the minimum wage together with fixed costs causing the shortage of part-time jobs.
According to the latest OECD report, Turkey is the least flexible labor market worldwide with
regards to employment. When labor markets are inflexible, for example due to the existence of
fixed costs, a minimum wage restriction makes it prohibitively expensive to offer part-time positions
to low-productivity workers. Most women, on the other hand, demand part-time work because of
their high value of home production. Facing the option of working full-time or not at all, they
choose to stay at home. This results in a high non-participation rate.
The rest of the paper is organized as follows. The next section describes the main argument in
length using both the theory and the underlying social and economic facts. Section 3 introduces the
econometric model. Section 4 gives description of the data and details of the estimation process.
Section 5 discusses some preliminary results. Conclusions and discussion of future extensions
follow.
2 Background
Over the last 50 years the proportion of part-time workers has increased significantly within the
labor forces of many countries, especially among women. However, in Turkey part-time jobs are
not as commonly observed. As shown in Figure 1, more than a quarter of working women are
holding part-time jobs in most countries. This rate is only 3 in 100 in Turkey (See Table B-1 in
Appendix B for a better look at the numbers).
Following Baslevent (2001), this paper claims that the lack of part-time jobs is an important
The mid 1950’s marked the beginning of the inflationary period in Turkey. Rising prices led to opposition from
unions to the local and unequal nature of the minimum wage and constant decline in its real value. In 1954, the law
was changed and minimum wage began to be controlled centrally, but it was not until 1967 that the minimum wage
was determined centrally by the national government.
Especially during the late 70’s and 80’s, minimum wage lost its effectiveness. Governments failed to adjust the
wage level against the inflation, which led to severe decline in real minimum wage level from time to time. It began
to be determined annually after 1982 in order to preserve the real minimum wage level but it was not quite enough.
This problem is being solved to some extent by adjustments made every six months for the last few years by the
minimum wage committee.
The minimum wage is now centrally determined and applied in all industries and in all cities. One criticized point
is that this application does not takes regional differences in living conditions into account. While minimum wage
can be enough for a 4-person family in some cities, in others it is not even enough to pay the rent (Istanbul is a good
example).
Minimum wage is lower for the workers who are under age of seventeen. It was 186.000 TL for seventeen-years-old
and older, 131.000 TL for sixteen-years-old and younger in 1988. About twenty percent of women in the sample
reports wages below net wages corresponding to net of this minimum wage level.
3
determinant of low participation rates for women, especially for married ones. Married women
have a higher-valued outside option since the division of labor in the household requires them to
be the main producer at home. When this is the case, a female of a given market productivity is
expected to supply fewer hours of labor than her male counterpart, or make the non-participation
decision more easily if she is forced to work long hours. The share of housewives among non-
participating women is strikingly high in Turkish data. In the 1991 round of the Household Labor
Force Survey (HLFS), 79 percent of women who do not participate in the labor force stated being
a housewife as the reason. This ratio was 80 percent in 1996. It is almost like being a housewife
is a job by itself. Household duties keep women at home when the labor market is not attractive
enough. As can also be seen in Figure 1, high female labor force participation goes hand in hand
with a high share of part-time jobs. This illustrates the fact that an increase in the number of
part-time positions makes market work attractive for more women.
particpation_rate
share_of_part_timers1.5 44
30
89
USA UK
Sweden
Denmark
GermanyFrance
Belgium
Ireland
Luxembou
Austria
Spain
Portugal
Italy
Poland
Turkey
Figure 1
It is a long established result in the literature that the hours-wage offers in the labor market
are bound together, and there is a wage premium for full-time employment (Oi, 1962; Barzel, 1973;
Moffitt, 1982). Due to productivity reasons or simply because of the fixed or quasi-fixed costs
associated with hiring, part-time jobs come with lower hourly wages. Women, while looking for
a job, may prefer flexibility [with regard to hours] over pay. As a result, part-time positions are
observed as low-pay, low-benefit jobs occupied by [married] women, in nearly all of the countries
4
on the above graph. However, in Turkey, existing part-time jobs exhibit different characteristics.
Average hourly wage in these part-time jobs almost doubles the average hourly wage in full-time
jobs. This observation surprises many scholars and some even claim that there is a wage premium to
part-time jobs in Turkey. This interesting phenomenon can be explained with the model introduced
in this paper. It can be shown that average part time wages are higher simply because there is
almost no part-time jobs at the low-pay end of the market due to the interaction of minimum wage
restrictions and high fixed costs associated with employment. The next two sections concentrate
on explaining how the minimum wage may be interacting with the economic environment in Turkey
and affecting the job market and, as a result, female employment.
2.1 Minimum wage, fixed costs and availability of part-time job
Consider an economy where there are costs associated with hiring that are equal to f hours of
production per worker for each workweek. Thus, each worker starts producing a surplus value
after the first f hours; this makes the average wage rate dependent on the hours worked. However,
minimum wage regulation does not take the existence of fixed costs into account and requires a
constant hourly wage independent of the length of the workweek. So, when there are fixed costs,
under a minimum wage regulation, no matter what the productivity level is, for each worker there
is an interval of hours where her per-hour-productivity is lower than the minimum wage. That is
why it becomes expensive for an employer to offer part-time jobs to workers who have productivity
levels lower than some threshold, βt, as depicted in Figure 2. In this case, only workers βa and βbwill be offered part-time jobs. Any worker with productivity level lower than βt (βc, βd and βT .
Here βT is the worker who works the highest possible number of hours) will be required to work
more than 40 hours2. Total production by the worker βT βa βb βt βc βd wmin f 40 T Hours of work
Figure 22 In this paper part-time work is defined as working less than 40 hours per week.
5
Total production by the worker βb βa wmin
f hb h* ha T Hours of work
Figure 3
Figure 3 illustrates two workers (worker a and worker b) with two different levels of constant
marginal productivity, βa and βb where βb > βa. Suppose that both of these workers desire to
supply the same number of hours to the market, which is equal to h∗. Before the imposition of aminimum wage (indicated with the bold dashed line) an employer can hire both workers, let them
work h∗ hours and pay them their average productivities as their hourly wages,βa(h
∗ − f)
h∗and
βb(h∗ − f)
h∗, respectively (with the market price of the good produced set equal to unity). However,
when there is a minimum wage, only the worker with productivity βb will be given this job. Worker
a, on the other hand, will be required to supply at least ha hours of work, since for any workweek
shorter than ha her average productivity is lower than the minimum hourly wage. Facing this
constraint on the hours of work, this worker may choose not to work at all.
The above discussion shows that a minimum wage can have negative employment effects when
there are high fixed costs and when the productivity distribution is highly left skewed. The
Turkish labor market can be characterized by both of these conditions: Low-median human capital
distribution and high fixed costs associated with employment3.
3According to OECD 2000 Statistics, Turkey has the least flexible labor laws worldwide regarding employment.
These inflexibilities reduce the productivity and the economy’s ability to adjust to changing technology. The OECD
report also states that flexibility of labor laws is positively correlated with a high possibility of finding jobs. In the
USA, the most flexible labor market according to the same study, 44% of the people who become unemployed are
able to find a job within a month. This ratio is “zero” in Turkey (OECD 2000).
Labor costs have increased significantly over the last few decades in Turkey. For example this increase was 19
percent between 1995 and 1999. In the same period labor costs decreased by 6.5 percent in Europe on average.
There are countries where the decrease is substantially higher ( In Ireland labor costs decreased by 26% between
6
2.2 Lack of part-time jobs and participation decision of married women
Suppose that on the supply side of the labor market, individuals maximize a well-behaved utility
function U = U(c, h ; Ai, i) choosing the amount of market work (h) they want to supply and the
level of a composite market good (c) they would like to consume. Their utility level for given c and
h depends on individual characteristics, Ai (gender, marital status, kids etc.)−indicators of homeproductivity or cost of participation− which are observable, and an unobservable preference shock,i− taste for market work− which is assumed to be normally distributed.As explained in the previous section, when there is a minimum wage in the market, the em-
ployer requires any given worker to work at least hmij (wmj , wi, f) hours. If hmij is lower than
h∗i (m, f,w, p;Ai, i), the worker’s decision process will not be altered. However, if h∗i < hmi she
can not work her desired hours anymore. At this point her participation decision will depend on
the difference between the utility level when working at the minimum required hours and utility
level at h = 0. For given levels of wage and non-labor income etc., this difference is a function of
preference parameters; only if it is greater than zero will she participate and supply hm hours to
the labor market at minimum wage, wm.
Figure 4 depicts two individuals with similar tastes for market work and the same home produc-
tivity measures but with different levels of market productivity, such as two women both married
with two kids, both with same non-labor income, one with low education (βk) and the other with
high education (βm)−assuming education is the only indicator of market productivity. When a
minimum wage is imposed, the low productivity worker −in this case, worker k− is going to berequired to work longer hours than she is willing to supply (hjk versus hmk ). Facing this, she may
choose not to work at all.
It is a well-recorded fact that men participate more and work more hours compared to women
and that single women participate more compared to married women. Figure 5 illustrates how
changes in home productivity (or cost of participation) or taste for work affect participation. This
picture shows two sets of workers with productivities the same as the ones in Figure 4. In this
these years).
Most of this observed increase in labor costs in Turkey is not due to the increases in the real wage levels. According
to the latest statistics, only 59 percent of the labor cost is the wage paid to the worker (TISK Journal). 41 percent
of the labor cost is the payments to the state, that is the social security payments. These payments constitute only
around 20 percent of the labor cost in the EU countries listed on Table 1.
Labor costs are not restricted to social security payments and wages. Besides the high costs of hiring there are costs
of firing. Most significant of these costs is the “seniority compensation” or severance pay. Any employee working
longer than a year in a institution, which has more than 10 employees, is entitled to severance that is equivalent to
the pay of 30 days per year for all years of his work. This pay is equal to at most 20 days in European countries
and it does not exist in most. For example, USA, Germany, Finland, Denmark, Ireland etc. do not have such a
regulation. The burden of the severance is reported to be around one-twelfth of Turkish GDP.
7
picture second set of workers have different preferences (indifference curves indexed with I) corre-
sponding to higher taste for work or lower home productivity (as in case of the men compared to
women). It can be seen that both individuals supply positive hours to market work under these
new preferences, since now the minimum hours constraint is not binding for either of them. Figure
5 demonstrates the main point about the employment effects of minimum wage and fixed costs:
they are going to be felt more severely by low-productivity individuals and by individuals with low
taste for market work who supply less hours - two sets of people whose intersection consists mostly
of women. Consumption Goods βm J1
βk J2 wmin Time hk
m hjm hjk hmm T-f T
Figure 4
Consumption Goods I1 βm J1
βk I2 J2 wmin Time hik hk
m hjm hjk hmm T-f T
Figure 5
8
3 Econometric Specifications
The main econometric difficulty in this analysis arises from the fact that I can not observe which
workers are at their lower bounds and which are working their desired hours. Moreover, I cannot
see which non-participants are constrained and would like to supply positive hours and which do
not. Therefore, I start by assuming that everybody has some level of desired hours, which, in this
model is represented as a linear function
h∗i = α0 + α1
µw
p
¶i
+ α2
µm− fw
p
¶i
+ α3Ai + 1i (1)
which implies
U = U(ci, hi; Ai, i) =
µh
α2− k
¶exp
µ(α0+α2ci + α3Ai + 1i)− k
h− k
¶as the utility function [where k = α1
α2f< h, and α2 < 0 − Slutsky Condition]4. In the latent
desired hours equation, m is non-labor income and p is the price of market goods [that is going to
be set equal to unity]. The labor supply choice is defined by two more latent indices in addition
to (1); offered wage (2) and minimum required hours5 (3), which is a function of the fixed costs,
minimum wage and the marginal productivity of the worker:
wi = Xiβ + i2 (2)
hmij =wi f
wi −wmj
(3)
where Xi are the individual productivity characteristics like education. Error terms 1 and 2 are
assumed to be independently and identically distributed as standard normals for the preliminary
analysis. This assumption can be relaxed for future analysis and possible heterogeneity in the
distribution of errors for different demographic groups can be considered.
As stated earlier, I do not observe either h∗ or hm. However, I know h, observed working hours,
if the individual is active in the labor market. Since, in this model, h is either desired hours or
minimum required hours, I can trace back to the conditions governing the participation decision,
4Pencavel worked out the utility function in his Handbook Chapter.
5Minimum required hours for each individual is found according to the following:
wminj = whi =wi(H
mini − f)
Hmini
9
and uncover the rules determining the choice of hours. Figure 6 illustrates the regions regarding
participation behavior in the plane of “desired” and “required minimum” hours.
As long as the individual desires longer workweeks than the minimum workweek that she is
offered, she is not going to be constrained by the minimum hours requirement and she is going to
work her desired hours. However, when desired length of workweek (given it is more than zero)
is shorter than the minimum one offered to her, she is going to be forced to choose between not
working and working the required minimum. She is going to work hm hours at minimum wage
only if it is more desirable to do so. That is,
hi = h∗i if T > h∗i > hmij , wmi < wi, and if h∗ > 0 (Region I)
hi = hmi if T > hmij > h∗i , h∗i > 0 , and if U(h = hm) > U(h = 0) (Region II)
hm (T,T)
(Region III) ( Region IV ) w>wm, h*>0 , hm>h* and U(h= hm)< U(h= 0) h=0
w>wm , h*<0 h=0 ( Region II ) w>wm, h*>0 , hm>h* ( Region I )
and U(h= hm)> U(h= 0) h=hm w>wm, h*>hm>0 h=h* f (0 , 0) h*
( Region V ) w < wm h = 0
Figure 6
10
According to this setting, there are three groups among the non-participants. The first group
is the group of individuals who would supply positive hours if they were not constrained. They
are demanded longer hours than they were willing to supply and facing this set of choices, they
prefer not to participate. For the second group of non-participants, on the other hand, the desired
workweek is less than or equal to zero. They are the ones who willingly choose not to participate.
The last group of non-workers consists of individuals who are undesirable in the market when there
is a minimum wage, that is, their productivity is lower than the minimum wage. In summary,
hi = 0 if hmi > h∗i , hmi > 0, h∗i > 0 and U(hi = hmi ) < U(hi = 0) (Region III)
or if h∗i < 0 and wmj < wi (Region IV )
or if wmj > wi (Region V )
Given these five events, the log likelihood function6 can be written as
L =TX
h>0
logQ+Xh=0
log q
where
Q =
hk(h| Region I , Xi, Ai, σ1, σ2, w
mj , mi)
iPr( Region I | Xi, Ai, σ1, σ2, w
mj , mi)
+hk(h| Region II, Xi, Ai, σ1, σ2, w
mj , mi)
iPr( Region II | Xi, Ai, σ1, σ2, w
mj , mi)
and
q = 1− Pr £h∗ > 0, hm > 0, U(h = hm) > U(h = 0) | Xi, Ai, σ1, σ2, wmj ,mi
¤=
ÃPr(U(h = hm) < U(h = 0), 0 < h∗ < hm < T | Xi, Ai, σ1, σ2, w
mj ,mi)
+Pr(wm > w | Xi, Ai, σ1, σ2, wmj ,mi) + Pr(h
∗ < 0 | Xi, Ai, σ1, σ2, wmj ,mi)
!
The model is identified by the non-linear structure of a generalized Tobit-type model. Variances
of the error terms are set equal to unity for simplicity in the preliminary analysis. However,
this simplification comes with the price of possibility of non-robustness since maximum likelihood
estimators are inconsistent if the underlying distributions are misspecified. Moreover, it is possible
that there are several local maxima and I am stuck in a “bad” one. Avoiding this possibility
requires a long and burdensome period of investigation.
6The derivation of the likelihood function is given in Appendix A.
11
4 Data
The data set used is from the Turkish Household Labor Force Supply Survey, which has been
conducted twice a year by the State Institute of Statistics of Turkey, from 1988 to 1999 and
quarterly since 2000. In total, about 14, 000 to 22, 000 households are surveyed each time, both
from rural and urban areas. This data set sheds light on the labor supply behavior of individuals
age 12 and older. The analysis here use the data from the October 1988 round of this survey since
it is the only one publicly available.
In the 1988 round, 102, 062 individuals residing in 22, 320 households nationwide are surveyed.
In this data set participation for women is around 17 percent in cities, very similar to the census
results. Participation rates, though, show great variation with education and marital status.
There are significant drops in participation rates as education falls below college level (71 percent
at college level and 9 percent for primary school graduates) and as women get married (37 percent
for singles, 11 percent for married). In the survey, nonworking women are asked if they would
like to work and the ratio of the who are ready to start working is higher among married and
low-educated (although slightly in some cases) suggesting that more of those women are the ones
who are unwillingly staying out of the market. And for the most non-working women the reason
for not working and not looking for a job is being a housewife which emphasis the importance of
home production.
For the preliminary analysis, I use a subsample of 5, 782 married women between ages 20 and
45 living in the cities with 400, 000 or more population. I exclude women younger than 20 from
the sample minimum wage requirements are different for some people in this age group. I do not
include women over 45 to avoid the large population of retirees. Around the year of this survey,
women had the option of retiring after 25 year of service. Women in the sample either did not
work the week preceding the interview or they were employed as wage and salary workers. I use
data only on women who are working at most one job and who are not currently enrolled in school,
either full time or part time. Table 1 gives the descriptive statistics.
Table 1: Descriptive Statistics
Variables # of obs. mean st.dev. min. max. median
Hours worked (if working) 465 39.43 8.15 9 74 40
# of children (ages 0-6) if have kids 2887 1.39 .62 1 4 1
# of children (ages 0-6) 5782 .69 .82 0 4 0
Education 5782 2.05 1.01 1 5 2
Age 5782 32.12 6.92 20 45 32
In this subsample the mean education is little over primary school. 81 % of the women
interviewed have 7 or less year of schooling (Last degree they have completed is primary school).
12
University graduates constitute 3 percent of the women and about 32 percent of the working women
in the subsample. The labor force participation rate for this subsample is 7.96 percent. These
women work 39.43 hours on average. 85.99 percent of working women work 40 hours or more and
only 5.31 percent work 20 hours or less (2.55 percent of women work between 25− 40 hours, 6.37percent work less than 25 hours).
Table 2 gives the distribution of hours in each sector. In almost all of the sectors mean hours
are higher than 40 hours. Service sector is the largest female employing sector and it is also the
only sector with average hours less than 40 hours per week.
Table 2: Hours & SectorsSectors # of obs mean st.dev. min max
finance 60 40.91 5.10 14 60
service 252 36.47 8.57 9 64
manufacturing 111 43.79 5.89 16 74
other 48 43.10 6.05 25 60
I use education(primary school, middle school, high school and college dummies), age, age
squared (Xi), number of young kids (ages 0-2 and 3-5), extended family dummy (Ai) and non-
labor income (m) as the explanatory variables in the preliminary analysis. There are few points
problematic about the data, for example, wages and the non-labor income measure are not directly
available. There is no record of asset income, etc. I use the per capita labor income in the family,
except woman’s wage income, as a proxy for her non-labor income. The problem I have with the
wages is more complicated. In the survey, individuals are asked their usual per week working hours,
and how much they worked last week. However, they report how much they earned the month
preceding the interview. I approximate the weekly labor income using these figures, making sure
that the individuals were working for the whole month for which they report the income. Three
observations which are not meeting this criteria are excluded from the sample used for the analysis.
The data set is cross-sectional and the nominal level of minimum wage is constant over the
country. I create variations in the minimum wage using the province level CPI (I used price
indexes supplied by the Central Bank of Turkey) and location adjustments. I keep the Ankara
prices as the base and divide the minimum wage in the other provinces with ratio of their prices to
the prices in the capital. Then I multiply this minimum wage measure with the location dummy
(which classifies the cities with respect to their population size). This measure should reflect the
differences in the real value of minimum wage across individuals even though they all face the same
nominal level. I made the same adjustment to non-labor income and wage measures, too.
13
5 Estimation
In this preliminary analysis, parameters of the wage function are estimated directly from the data
[corrected for selection bias], and they are set fixed for the estimation of other parameters of the log
likelihood function. Starting values for other parameters are obtained from the data as well. I set
a value for the fixed cost and estimated desired hours equation using OLS. I used OLS estimates
of the parameters as starting values for the maximization problem, together with the fixed value
of f which I used to estimate those values. Maximization procedure fails to calculate standard
errors for these estimates, so they are approximated using the sum of outer product of the scores
for each observation, a method proposed by Berndt, Hall, Hall, and Hausman in 1974. Here the
asymptotic standard errors are square roots of the diagonal elements of the following matrix:
"NXi=1
si(θ)si(θ)0#−1
where N is the number of observations, and θ is the vector of estimated parameters.
Estimates suggests that existence of young kids in the household decreases the number of
hours willingly supplied to the labor market. Desired hours decreases also by net non-labor
income, not a very surprising result. Moreover, women from extended families tend to supply less
hours. Considering that these sort of families are usually observed in more conservative segments
of society, this results supports the idea that traditional role of woman is to be the homemaker,
not a breadwinner for the family.
Wage equation estimates suggest positive returns to education especially at college level. Col-
lege graduate women earn significantly more compared to illiterates. Return to college education
is four times higher than returns to high school degree. This partially explains big gaps between
participation rates across different education levels.
Estimated participation rate is 8.51 percent. None of the participants are restricted, that is,
estimates suggest that all working women are working their desired hours. However, 35.32 percent
of all women are restricted in the sense that they want to supply positive hours but either not
desired as workers or constrained by high required minimum hours. Conditional on being a non-
participant, 32.45 percent of women want to work and they are welcome in the market but they are
asked to work more hours than they are willing to, and 6.049 percent of women are not offered any
job although they wish to supply positive hours to the market. Table 3 reports the preliminary
estimation results. Tables 4a-g summarize the participation probabilities these estimates suggest
and others which are generated via simulations under different minimum wage and fixed cost levels.
Table 5 gives the hours distribution for the estimates and the simulated events.
14
Table 3: Coefficient Estimates of Desired Hours,
and Wage Equations
Desired Hours Wage
constant2.3306
(std)
5.9708
(std)
primary school --0.0137
(std)
secondary school -0.4198
(std)
high school -0.6687
(std)
college -2.0527
(std)
kids 0-2-0.0861
(std)-
kids 3-5-0.4440
(std)-
age-0.5930
(std)
0.0594
(std)
age squared --0.0065
(std)
extended family-0.4806
(std)-
wage0.0169
(std)-
net non-labor income-0.7123
(std)-
fixed cost9.2928
(std)-
value of log likelihood function -1542.64
**Standard deviations will be reported soon.
15
Table 4-a f=fdata,wmin=wmindata
Event Probability
H>0 0.0851
H=0, Hmin>0, H∗>0 0.2969
Hmin>0, H∗>0 | H=0 0.3245
H=0, Hmin<0, H∗>0 0.0505
Hmin<0, H∗>0 | H=0 0.0649
H=0, Hmin>0, H∗<0 0.3189
Hmin>0, H∗<0 | H=0 0.3485
H∗>0 0.4374
Table 4-b f=fdata,wmin=wmin
data*0.90
Event Probability
H>0 0.1057
H=0, Hmin>0, H∗>0 0.3087
Hmin>0, H∗>0 | H=0 0.3452
H=0, Hmin<0, H∗>0 0.0243
Hmin<0, H∗>0 | H=0 0.0273
H=0, Hmin>0, H∗<0 0.3229
Hmin>0, H∗<0 | H=0 0.3611
H∗>0 0.4388
Table 4-c f=fdata,wmin=wmin
data*0.80
Event Probability
H>0 0.1228
H=0, Hmin>0, H∗>0 0.3091
Hmin>0, H∗>0 | H=0 0.3523
H=0, Hmin<0, H∗>0 0.0069
Hmin<0, H∗>0 | H=0 0.0078
H=0, Hmin>0, H∗<0 0.3400
Hmin>0, H∗<0 | H=0 0.3876
H∗>0 0.4388
Table 4-d f=fdata,wmin=wmin
data*0.10
Event Probability
H>0 0.2496
H=0, Hmin>0, H∗>0 0.1892
Hmin>0, H∗>0 | H=0 0.2521
H=0, Hmin<0, H∗>0 0.0000
Hmin<0, H∗>0 | H=0 0.0000
H=0, Hmin>0, H∗<0 0.5336
Hmin>0, H∗<0 | H=0 0.7110
H∗>0 0.4388
Table 4-e f=fdata,wmin=wmin
data*0.01
Event Probability
H>0 0.2641
H=0, Hmin>0, H∗>0 0.1747
Hmin>0, H∗>0 | H=0 0.2374
H=0, Hmin<0, H∗>0 0.0000
Hmin<0, H∗>0 | H=0 0.0000
H=0, Hmin>0, H∗<0 0.5581
Hmin>0, H∗<0 | H=0 0.7584
H∗>0 0.4388
16
Table 4-f f=fdata-1,wmin=wmin
data
Event Probability
H>0 0.0834
H=0, Hmin>0, H∗>0 0.2992
Hmin>0, H∗>0 | H=0 0.3264
H=0, Hmin<0, H∗>0 0.0322
Hmin<0, H∗>0 | H=0 0.0351
H=0, Hmin>0, H∗<0 0.3213
Hmin>0, H∗<0 | H=0 0.3506
H∗>0 0.4147
Table 4-g f=fdata-2,wmin=wmin
data
Event Probability
H>0 0.0752
H=0, Hmin>0, H∗>0 0.2989
Hmin>0, H∗>0 | H=0 0.3231
H=0, Hmin<0, H∗>0 0.0159
Hmin<0, H∗>0 | H=0 0.0174
H=0, Hmin>0, H∗<0 0.3298
Hmin>0, H∗<0 | H=0 0.3600
H∗>0 0.4388
Table 5 Distribution of Hours
mean std min max median
estimated 30.55 5.57 21 51
wmin=wmindata*0.90 29.20 5.71 21 53
wmin=wmindata*0.80 28.11 5.96 19 53
wmin=wmindata*0.10 22.21 7.38 12 53
wmin=wmindata*0.01 21.63 7.56 11 53
f=fdata-1 27.07 4.93 20 48
f=fdata-2 23.86 4.41 18 42
Decreases in minimum wage have the expected effect on participation: ratio of women working
increases gradually as minimum wage decreases, up to 26 percent at 99 percent decrease(Table
4-b 10%, 4-c 20%, 4-d 90% and 4-e 99% decline in minimum wage). Moreover, distribution
of hours worked changes; means and minimums decrease as minimum wage decreases, indicating
more women working at the low end of hours distribution.
Table 4-f shows the probabilities when fixed cost is decreased by the value of one hour worth of
production and Table 4-g reports the same probabilities in case of two hours of production worth of
decline in fixed costs. Decreases in fixed costs, not only reduce the mean hours of work and range
of hours observed, but also narrow the extensive margins of participation. Income effect arising
from the impact of decline in fixed costs on net non-labor income must reduce the desired hours
for women more than it reduces the required minimum resulting in even less number of working
women. This shows that at least at low levels of change, reducing minimum wage is a better
strategy to promote female labor force participation.
Using the estimates I construct the desired hours and minimum hours for each individual and
generate the implied working hours distribution (Figure 8). Comparing this distribution with the
17
distribution of actual hours (Figure 7) in the data shows that the specifications of the model are
not very successful. This may be indicating the need for the re-specification of the model.
020
4060
Percent
0 20 40 60 80h
Figure 7
05
1015
Percent
20 30 40 50h_est
Figure 8
18
There are other exercises that can be done with the estimates. For example, I can identify
the intensive and extensive effects of changes in the individual characteristics on the participation
decision. The effect of any change in the level of constrained hours is going to be in two ways:
changes in the hours worked and changes in the participation decision. To be able to analyze
these dynamics I will use the estimated labor supply behavior. For this model, unconditional
expectation of labor supply can be written in two parts; expected value of hours of work given it is
the minimum required hours and expected value of hours of work given it is the desired level, that
is:E [hi] = E [hi|h∗i > hmi , wi > wm] P ( T > h∗ > hm, w > wm)
+
"E [hi|hmi > h∗i > 0, U(hi = hmi ) > U(hi = hmi )]
Pr( 0 < h∗ < hm < T, U(h = hm) > U(h = 0))
#Then the marginal change in the unconditional labor supply due to a marginal change in a given
variable Y (Xi , Ai) can be written as
∂E [hi]
∂Y=
∂Ehhi|h∗i > hmi , wi > wm
j
i∂Y
P ( T > h∗ > hm, wi > wmj )
+
µE£hi|h∗i > hmi , wi > wm
j
¤ ∂P ( T > h∗i > hmi , wi > wmj )
∂Y
¶
+
∂E [hi|hmi > h∗i > 0, U(hi = hmi ) > U(hi = hmi )]
∂YPr( 0 < h∗i < hmi < T, U(hi = hmi ) > U(hi = 0))
+
E [hi|hmi > h∗i > 0, U(hi = hmi ) > U(hi = hmi )]∂ Pr( 0 < h∗i < hmi < T, U(hi = hmi ) > U(hi = 0))
∂Y
This expression decomposes the total changes in the expected hours into four parts: (1) changes in
the hours for those who are working their desired hours conditional on being able to work desired
workweek; (2) the change in the probability of being able to work desired hours weighted by the
expected value of hours for who work desired hours; (3) changes in the hours for those who are
working imposed required minimum workweeks conditional on being restricted; and (4) the change
in the probability of being required to work minimum hours weighted by the expected value of
hours for the constrained workers.
6 Conclusion & Future Extensions
This paper provides a discussion of the possible forces behind the low (and still decreasing) female
labor force participation in Turkey. It attempts to explain this decline by pointing out the possible
negative employment effects of minimum wage regulations in inflexible markets. It is shown that
19
the minimum wage, together with fixed costs, may leads employers to put constraints on the length
of workweeks they offer. The model predicts that household responsibilities (number of children
etc.) decrease while education (via wages) increases the hours supplied by married women. For
given demographics, increases in the level of the minimum wage and the fixed cost increase the
probability of being constrained in the labor market. On the other hand, the more educated a
woman is, the less constrained she will be. This implication of the model can be an explanation
for the differences in the characteristics of the part-time jobs in Turkey compared to the other
countries.
Next step in the project is to complete the empirical analysis for the basic form of the model.
Once I get reasonable estimates for the simplest form, I can improve the model. There are some
things I would like to re-specify for the future versions of this paper. With this setting, there
is no place for a non-monotonic relationship between hours supplied and fixed costs. Also there
is no room for non-linear responses to wages. I should redo the analysis with a more flexible
specification. I tried to keep the number of parameters as low as possible for this initial analysis.
For example, although they are crucial in my setting, possible heterogeneity of fixed costs is ignored
in the preliminary empirical analysis for the sake of simplicity. Province-level variables should be
included in the estimations as indicators of the market conditions effecting the non-wage labor
costs. Size of the service sector in the region and share of large enterprises (+10 employees)
are additional variables which can be considered. On the supply side, variables like size of the
household or number of elderly in the household may be included in the estimation. Furthermore,
the non-linear character of the household production can be considered.
The data set can be enriched by inclusion of single females and maybe males. This will make
it possible to include marital status and gender as explanatory variables. Additional variables
will increase the strength of the approximation. However, the available data set lacks potentially
significant information, the inclusion of which may increase the explanatory power of the model.
Re-estimation of the model with a data set which reports the hours of work over a longer period
of time, and reports a greater variety of individual variables such as non-labor income, education
of parents, health status indicators, etc. can be considered. This data set only asks about
earnings and hours worked in the week preceding the survey. This may not be a true indicator of
participation behavior. This is a cross-sectional data and variances in the minimum wages are only
through price and location adjustments. For a better understanding of the effects of minimum
wage restrictions, it would be helpful if I were able to follow the sample over time as real minimum
wage changes. Moreover, other sources of hours dependency of wages can be considered for the
analysis. When possible nonlinear structures of marginal productivity considered demand side can
be more effectively studied.
20
References
[1] Barzel, Yoram, (1973), “The Determination of Daily Hours and Wages”, Quarterly Journal of
Economics, 87(2), pp.220-238
[2] Baslevent, C.(2001) Essays on Female Labor Supply in Turkey , Ph.D. Thesis, Department of
Economics, Bogazici University.
[3] Dayioglu, M. (1998), “Labor Market Participation of Women in Turkey” in F.
Acar and A. Ayata (eds.), Women’s Identities and Roles in the Course of Change:
Central Asia, East and Central Europe and Turkey, Duke University Press, forthcoming.
[4] Ecevit, Y. (1990), “Kentsel Uretim Surecinde Kadin Emeginin Konumu ve Degisen Bicimleri”
in S.Tekeli (ed), Kadinin Bakis Acisindan 1980’ler Turkiye’sinde Kadin, Istanbul: Iletisim
Yayincilik.
[5] Mincer, J. (1985) “Intercountry Comparisons of Labor Force Trends and of Related Develop-
ments: An Overview” Journal of Labor Economics, 3(1), pp. S1-S32.
[6] Moffitt, R. (1984) “The Estimation of a Joint Wage-Hours Labor Supply Model”, Journal of
Labor Economics, 2(4), pp.550-66.
[7] OECD 2000 Country Report, OECD Webpage (www.oecd.org)
[8] Oi, Walter (1962), “Labor as a Quasi-Fixed Factor”, Journal of Political Economy, 70, pp.
538-555.
[9] Özar, S.(1996), “The Employment Aspects of the Informal Sector” in Tuncer Bulutay (ed.),
Informal Sector (II), pp.175-203. Ankara: State Institute of Statistics, Printing Division.
[10] Pencavel, J. “Labor Supply of Men: A Survey,” in O. Ashenfelter and R. Layard (eds.),
Handbook of Labor Economics, Vol. 1. North Holland, 1986.
[11] Schultz, T. P. (1990) “Women’s Changing Participation in the Labor Force: A World Perspec-
tive”, Economic Development and Cultural Change, 38 (3), pp. 719-42.
[12] Shorter, Fred (1995), “The Crisis of Population Knowledge in Turkey.” New Perspectives on
Turkey, no.12 (Spring), pp.1-31.
[13] SIS (State Institude of Statistics) Webpage, HLFS Database
[14] TISK Journal, TISK(Turkish Confederation of Employer Associations) Webpage
(www.tisk.org.tr)
21
[15] Tunali, I. (1997) “To Work or not to Work: An Examination of Female Labor Force Par-
ticipation Rates in Urban Turkey” Proceedings of the 4th Annual ERF Conference held in
Beirut.
[16] Tunali, I. and C. Baslevent (2000) “Estimation of Female Labor Supply Parameters When
Self-Employment is an Option” Paper presented at the 7th Annual ERF Conference held in
Amman.
[17] Tuncay, A., Türk Sosyal Güvenlik Sisteminde Yeniden Yapilanma, TÜSIAD: Istanbul,1997.
AppendicesAppendix-A
Derivation of the Likelihood FunctionThe individual’s problem is to
maxc,h
µα2hi − α1
α22
¶exp
µα2 (α0 + α2(m+ wi(hi − γf)) + α3Ai + 1i)− α1
α2hi − α1
¶s.t. pc = mi + wi(hi − γf)
where m is nonlabor income and γ is an index variable which is equal to one if h > 0 and zero
otherwise. This optimization problem gives
h∗ = α0 + α1
µw
p
¶i
+ α2
µm− fw
p
¶ij
+ α3Ai + 1i
as the desired hours equation. This model has two more latent indexes:
wi = Xiβ + i2
hmi =wif
wi −wmj
Then for a worker
hi = h∗i (works desired hours) if
h∗i > hmi and wi > wmj
hi = hmi (works required minimum hours ) if
0 < h∗i < hmi and U(hi = hmi ) > U(hi = 0)
22
hi = 0 (desires to work but is restricted) if
0 < h∗i < hmi and U(hi = hmi ) < U(hi = 0)
hi = 0 (does not want to work) if
h∗i < 0 and wmj < wi
and
hi = 0 (can not work) if
wmj > wi
Then the log-likelihood function is:
logL =Xh>0
logQ+Xh=0
log q
where
Q =
hk(h| Region I , Xi, Ai, σ1, σ2, w
mj , mi)
iPr( Region I | Xi, Ai, σ1, σ2, w
mj , mi)
+hk(h| Region II, Xi, Ai, σ1, σ2, w
mj , mi)
iPr( Region II | Xi, Ai, σ1, σ2, w
mj , mi)
and
q =
ÃPr(Region III | Xi, Ai, σ1, σ2, w
mj , mi) + Pr(Region IV | Xi, Ai, σ1, σ2, w
mj ,mi)
+Pr(Region V | Xi, Ai, σ1, σ2, wmj , mi)
!
[k(.) is the conditional probability density function of the hours of work variable given dependent
variables, nonlabor income and minimum wage levels and the preference coefficients].
Derivation of k(h| Region I , Xi, Ai, σ1, σ2, wmj , mi) is straigthforward. The latter expression,
k(h|Region II, Xi, Ai, σ1, σ2, wmj , mi) can be derived as follows:
k(h|Region II,Xi, Ai, σ1, σ2, wmj ,mi) =
∂K(h|Region II,Xi, Ai, σ1, σ2, wmj ,mi)
∂h
where K(.) is the conditional cummulative density function of hours of work variable given depen-
dent variables, nonlabor income and minimum wage levels and the preference coefficients which
23
can be explicitly expressed as follows:
K(h|0 < h∗ < hm, U(h = hm) > U(h = 0), Xi, Ai, σ1, σ2, wmj , mi)
= Pr(h|0 < h∗ < hm, U(h = hm) > U(h = 0), Xi, Ai, σ1, σ2, wmj , mi)
=Pr(0<h∗<hm<h, U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, w
mj , mi)
Pr(0<h∗<hm, U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, wmj , mi)
=Pr(0<h∗<hm, U(h=hm)>U(h=0) | hm<h, Xi, Ai, σ1, σ2, w
mj , mi) Pr(h
m<h| Xi, Ai, σ1, σ2, wmj , mi)
Pr(0<h∗<hm,U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, wmj , mi)
=Pr(0<h∗<hm, U(h=hm)>U(h=0) | hm<h, Xi, Ai, σ1, σ2, w
mj , mi) Pr
µ(Xiβ+ i2)f
(Xiβ+ i2)−wmj <h
¶Pr(0<h∗<hm,U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, wmj , mi)
=Pr(0<h∗<hm, U(h=hm)>U(h=0) | hm<h, Xi, Ai, σ1, σ2, wmj , mi) Pr
Ãi2<
(h−f)Xiβ−wmj h
(f−h)
!Pr(0<h∗<hm,U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, wmj , mi)
=Pr(0<h∗<hm, U(h=hm)>U(h=0) | hm<h, Xi, Ai, σ1, σ2, wmj , mi) Φ
Ã(h−f)Xiβ−wmj h
(f−h) σ2
!Pr(0<h∗<hm,U(h=hm)>U(h=0) | Xi, Ai, σ1, σ2, wmj , mi)
Thus, k(.) is equal to
Pr(0<h∗<hm,U(h=hm)>U(h=0) | hm<h, Xi,Ai,σ1,σ2,wmj ,mi) (−wmj f
(f−h)2σ2)φ(
(h−f)Xiβ−wmj h
(f−h) σ2 )
Pr(0<h∗<hm,U(h=hm)>U(h=0) | Xi,Ai,σ1,σ2,wmj ,mi)
And7
q =
Pr
(α2h
mi − α1) exp
µα2(α0+α2m+α2hmi wm+α3Ai+ 1i)−α1
α2hmi −α1
¶<
(−α1) exp³α2(α0+α2m+α3Ai+ 1i)−α1
−α1´
, h∗ > 0,w > wm | Xi, Ai, σ1, σ2, w
mj ,mi
+Pr
hh∗ < 0, w > wm | Xi, Ai, σ1, σ2, w
mj ,mi
i+Pr
hw < wm | Xi, Ai, σ1, σ2, w
mj ,mi
i
7Few components of this expression do not have analytical solutions. These probabilities are calculated using
numerical methods.
24
=
Pr
(α2h
mi − α1) exp
µα2(α0+α2m+α2hmi wm+α3Ai+ 1i)−α1
α2hmi −α1
¶<
(−α1) exp³α2(α0+α2m+α3Ai+ 1i)−α1
−α1´
, Xβ − wm > − 2i,
α0 + (α1 − α2f)Xβ + α2mi + α3Ai > 1i + 2i(α1 − α2f)
+Pr
hα0 + α1
³Xβ + i2
´i+ α2
³m− f(Xβ + i2)
´i+ α3Ai + 1i < 0, Xβ + i2 > wm
i+Pr
hXβ + i2 < wm
i
=
Pr
(α2h
mi − α1) exp
µα2(α0+α2m+α2hmi wm+α3Ai+ 1i)−α1
α2hmi −α1
¶<
(−α1) exp³α2(α0+α2m+α3Ai+ 1i)−α1
−α1´
, Xβ − wm > − 2i,
α0 + (α1 − α2f)Xβ + α2mi + α3Ai > 1i + 2i(α1 − α2f)
+Φ
·−[α0+(α1−α2f)Xβ+α2m+α3Ai]√
σ21+(α1−α2f)σ22, −Xβ+wm
σ2
¸+Φ
hXβ−wm
σ2
i
25
Appendix-B
Table B-1: Female Participation and Part-time Jobs in USA, European
Countries and Turkey, 1991
Female Participation Female Part-timers as
percent of Working-age Population percent of Female Employment
USA 69 25
Netherlands* 53 66
UK 69 44
Sweden 89 41
Denmark 79 34
Germany 63 33
France 65 28
Belgium 53 28
Ireland 38 21
Luxembourg 58 20
Austria 55 18
Spain 41 15
Portugal 63 12
Italy 45 12
Finland* 77 11
Poland 52 13
Turkey 30 3Sources: EU Labor Force Survey Data, 1994 ; Employment in Europe,
1995;SIS—Turkey, Women Indicators and Statistics and SIS HLFS 1994
Definition of part-time is different in US and Turkish data. In the US,
anybody working less than 35 hours per week is considered part time
worker, rather than people working part time in their main jobs.
In Turkey, it is anybody working less than 40 hours.
*Finland and Netherlands are excluded from the graph
26