Transcript 1-Factors Affecting Employment - De Jesus

FACTORS AFFECTING EMPLOYMENT
OUTCOMES OF FILIPINO JOBSEEKERS:
A DURATION ANALYSIS OF UNEMPLOYMENT
IN THE PHILIPPINES
By
Claire Dennis S. Mapa and Jeremy L. De Jesus
Presented by
Jeremy L. De Jesus
Bangko Sentral ng Pilipinas
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Factors Affecting Employment
Outcomes of Filipino Jobseekers:
A Duration Analysis of Unemployment
in the Philippines
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
BACKGROUND
Unemployment has been stubborn
for years.
Unemployment
10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
2,728
6.6
2006
2007
2008
2009
2010
Number (in thousands)
Source: 2015 Yearbook of Labor Statistics, PSA
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
2011
2012
2013
Rate (in percent)
2014
3,000
2,900
2,800
2,700
2,600
2,500
2,400
2,300
2,200
2,100
2,000
BACKGROUND
2014 Unemployment Snapshot
As % of total unemployed
63.6
Male
49.1
Young*
35.3
College educated**
0.0
20.0
40.0
60.0
80.0
Source: 2014 Annual Labor and Employment Status, PSA
Notes: *aged 15-24 **college undergrads and grads
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
100.0
MOTIVATION
Informal or vulnerable employment
is the bigger problem but…
30.0
Poverty Incidence Among Population: 2012
(in percent)
29.0
25.2
21.9
20.0
18.7
10.0
0.0
Total
Employed
Vulnerable
Employed*
Unemployed
Source: 2012 Official Poverty Statistics for the Basic Sectors, PSA
Note: *self-employed and unpaid family workers
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
MOTIVATION
…economic and non-economic costs of
unemployment cannot be downplayed.
• Economic costs
– Foregone economic
output
– Reduction in
government revenue
– Decline in household
incomes and wealth
• Non-economic costs
– Adverse effect on
health
– Strained family
relations
– Rise in cases of
alcohol and substance
abuse
– Higher suicide and
crime rates
– Negative impact on
human capital
The longer the unemployment spell, the worse the effects become.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
MOTIVATION
Aggregate measures of unemployment flows
for macroeconomic forecasting are available…
1.00
0.75
Outflow Rate* by Unemployment Duration
(in percent)
0.77
0.69
0.54
0.50
0.36
0.25
0.00
d<1 mo.
d<3 mos.
d<6 mos.
2009
9th
d<12 mos.
2010
Source: Key Indicators of Labor Market
edition, ILO (Estimates based on Shimer, 2012)
Notes: *Refers to the monthly instantaneous rate of transition from
unemployment to employment; ‘No duration dependence’ rejected
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
MOTIVATION
…but policy analysis of unemployment
duration on a micro-level is lacking.
• Studies on unemployment duration in developing
countries is very limited (Dendir, 2007; Tansel and
Tasci, 2010).
• Although anecdotal evidence abounds, to date,
there had been no previous research in the
Philippines specifically on unemployment spell.
• The bulk of the micro-level policy research on
unemployment in the Philippines were based on
stock variables (e.g. unemployment rate, labor force
participation rate).
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
MOTIVATION
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
OBJECTIVES
What individual, household, and
community characteristics lead
to longer unemployment spells?
• This paper aims to examine how demographic and
socio-economic characteristics of active Filipino jobseekers affect the length of their unemployment
spell and the likelihood of exiting to employment.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
OBJECTIVES
Does unemployment beget
unemployment?
• Equally important, this study intends to determine if
the amount of time individuals already spent without
a job hurts their chances of exiting to employment.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Theory of Sequential Job Search
• Given that imperfect information exists in the labor
market, search is inevitable.
• Stigler (1962) proposed the fixed sampling approach
to job search i.e., choose a sample of n jobs at a cost
c per wage sampled and then accept the highest offer.
• McCall (1970) and Mortensen (1970) developed the
sequential “stopping” approach to job search i.e., jobs
are sampled one at a time and deciding on the sample
obtained to date whether or not to stop the search or
to continue.
• The sequential approach was deemed more
reasonable given that “search” takes place in real time
and that offers must be accepted shortly after they are
made.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Modern Job Search Framework
• In a stationary environment, the search for job will
continue until a job offer is received after which the
choice of whether to accept or reject the job offer will
be decided by comparing the wage offer with the
reservation wage (or the lowest wage at which a jobseeker will accept a job).
• If the offered wage is greater than or equal to the
reservation wage, the search will cease. Conversely,
if the wage offer is less than the reservation wage,
the search will resume.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Reservation Wage
where z is the net instantaneous income in
unemployment*, λ is the arrival rate of job offers, r is
the discount rate, q is the rate of job loss, w is the
offered wage.
*unemployment benefits less search cost
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Hazard Rate from Unemployment
Since a job-seeker becomes employed when he or
she receives a job offer which occurs at rate λ and
the offered wage is at the very least equal to his or
her reservation wage which occurs with probability
[1-F(wR)], the hazard rate from unemployment is
given as:
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Average Unemployment Duration
Assuming that the hazard rate is constant, the
probability that a job-seeker is still unemployed after
a spell of length t is:
Consequently, the average unemployment duration
Tu can be derived as follows:
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
FRAMEWORK
Search-theoretic Model
• The search-theoretic framework postulates that the
exit rate from unemployment will be a function of
variables that affect the probability of receiving a
job offer and variables that affect the probability of
accepting the job offer.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
METHODOLOGY
Data Description
(1) July 2009 and January
2010 rounds of the
Labor Force Survey
(LFS)
(2) 2009 Family Income
and Expenditure Survey
(FIES)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Sample rotation scheme
Year
Quarter
January
April
2007
July
October
January
April
2008
July
October
January
2009
April
(FIES year)
July
October
January
April
2010
July
October
January
April
2011
July
October
Adapted from: Barcenas (2004)
18
Rotation Cluster/Group
A1
B1
A2
B2
A3
B3
A4
B4
A1
B5
A2
B6
A3
B7
A4
B8
A5
B5
A6
B6
A7
B7
A8
B8
A7
B7
A6
B9
A5
B10
A8
B11
A9
B12
A10
B9
A11
B10
A12
B11
METHODOLOGY
Unit of Analysis
(1) Sample consists of individuals aged 15 to 64 who
were unemployed and were actively looking for
work in July 2009 (N=2,896).
(2) Taking out household-based attrition, the final
sample size of the person data is 2,734 (67.5
percent of which are right-censored observations
i.e., still unemployed, left the labor force, employed
abroad, and observed in July 2009 only).
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
METHODOLOGY
Duration Data from LFS
July 2009
January 2010
TIMELINE
Transition
Interval-censored
U-E :
Completed
U-U :
?
U-OLF :
?
?
U-OW :
?
U-A :
26 weeks
No. of weeks
looking for work, t0
Duration
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Rightcensored
METHODOLOGY
Variables on Receiving Job Offer
Personal
Characteristics
• Age
• Sex
• Marital Status
• Education
Local Labor Demand
• Unemployment Rate
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Previous Work
Experience
• First time to look for
work
Trainings (proxy)
• Registration in
public/private
employment agency
METHODOLOGY
Variables on Accepting Job Offer
Reservation Wage
(proxy)
• Minimum Wage
Schooling Status
• Attending School
Household Composition
• Dependents
• Non-dependents
• Informal Worker
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Unemployment Income
• Assistance from Abroad
• Assistance from Local
• Loans from Other
Households
• Withdrawal from
Savings
METHODOLOGY
Duration Model: Single-Risk Discretetime Proportional Hazards Model with
Flexible Baseline Hazard Specification
where θij is the probability that individual i has left
unemployment during interval j, Dij is a vector of
functions of the cumulative duration by interval j with
coefficients α, and xij is a vector of covariates with
coefficients β.
References: Prentice and Gloeckler (1978) and Jenkins (2005)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
METHODOLOGY
Piece-wise constant specification of
the baseline hazard
The baseline hazard αDij , is specified to be a step
function,
where D1,…D4 are dummies for time intervals
j=1,…4 which are as follows: 1-27 weeks, 28-54
weeks, 55-81 weeks, and 82 weeks and over.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
METHODOLOGY
Estimation Procedure
- For this study, the complementary log-log model was
selected as the estimation procedure due to the following
reasons:
(1) at low hazard values, the logistic and complementary loglog functions are virtually identical; and
(2) it builds in a proportional hazards assumption similar to the
Cox regression model where the estimated parameters (i.e.,
exponentiated coefficients) can be interpreted as hazard
ratios.
- In order to estimate the grouped data hazards model, the
person data was transformed to person-period data which
increases the final sample size to 4,658 obs with 890 failures
(i.e., total no. of exit-to-employment events).
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
RESULTS
What individual, household, and
community characteristics lead
to longer unemployment spells?
Holding other personal characteristics constant…
Age
• The hazard of exiting unemployment initially
increases as job-seekers become older but begins
to decline at about the age of 35-36.
Sex
• Females, in general, were estimated to face 0.867
of the hazard of males or they have 13.3% smaller
hazard than males. Note that a hazard ratio of less
than 1 indicates that exiting from unemployment is
occurring slower for females than for males.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
RESULTS
Holding other personal characteristics constant…
Marital Status
• Ever married job-seekers, in general, were estimated to face
1.205 of the hazard of never-married unemployed (or they
have 20.5% higher hazard than the never-married).
Sex and Marital Status (Interaction)
• Married women were estimated to face 0.866 of the hazard of
married men (or they have 13.4% smaller hazard than
married men).
Education
• College undergrads and college grads face 0.710 and 0.647
of the hazard of primary grads or 29% and 35.3% smaller
hazard than the reference group, respectively.
• Graduates of engineering and services programs face 0.535
and 0.342 of the hazard of primary grads or 46.5% and 65.8%
smaller hazard than the reference group, respectively.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
RESULTS
Controlling for personal characteristics and other factors…
Previous Work Experience
• New entrants to the labor force were estimated to face
0.674 of the hazard of experienced job-seekers (or they
have 32.6% smaller hazard compared to the reference
group).
Local labor demand
• For every 1% increase in unemployment rate, the
hazard to employment decreases by 2.7%.
Trainings (proxy)
• Job-seekers who registered in public employment
agency were estimated to face 1.329 of the hazard of
job-seekers who either approached employer directly,
friends or relatives, placed or answered advertisements,
or did some other methods of finding work (or they have
32.9% higher hazard than the reference group).
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
RESULTS
Controlling for personal characteristics and other factors…
Reservation wage (proxy)
• For every 10-peso increase in minimum wage, the hazard to
employment decreases by about 2%.
Household composition
• For every additional employed member in the household, the
hazard to employment falls by 9.9%.
• Jobseekers living with an informal worker in the household
have 13.4% higher hazard to employment than that of
jobseekers in households with no informal worker
Unemployment income
• As the amount of assistance received from abroad and
domestic sources increases by 10%, the hazard to employment
falls by 35% and 20%, respectively.
• As the amount of loans from other families increases by 10%,
the hazard to employment rises by 26%.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
RESULTS
Baseline Hazard
Shape of Baseline Hazard Controlled for Covariates
(with quadratic trendline superimposed)
1.40
1.20
Hazard
1.00
0.80
0.60
0.40
0.20
0.00
1-27 weeks
28-54 weeks
55-81 weeks
Intervals of Unemployment Duration
Source: Author’s computation
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
>=82 weeks
RESULTS
Does unemployment beget
unemployment?
• Yes! There is evidence indicating negative duration
dependence after 28-54 weeks. The hazard to
employment falls monotonically thereafter.
• The initial positive duration dependence indicates
that it is difficult to find and start a new job which is
common among those individuals who are first-time
job-seekers and have no previous work experience.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
REMARKS
Conclusions
• Several personal,
household, and
community attributes
were identified as
influential factors to
unemployment duration.
• In the medium- and longterm, unemployment
begets unemployment.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
REMARKS
Policy Implications
• Preparing our youth for their entry to the labor force is
desirable.
• Giving employment opportunities to married women may
improve the labor force participation of women in
general.
• It is beneficial to match the skills of the workforce with
the demands of the labor market.
• The disincentive albeit unintentional effect of interhousehold transfers on labor supply decisions of leftbehind families of migrant workers is a cause for
concern.
• A review of the LFS current design and questionnaire
should be seriously considered by the PSA.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City