Job Prospects and Internal Migration in China

Outline Introduction Research Gaps & Objectives Theoretical Model Data & Empirical Analysis Main Results Robustness Check Concluding Remarks
Job Prospects and Internal Migration in
China
Huaxin Wang-Lu, Octasiano M. Valerio Mendoza
IQS School of Management, Universitat Ramon Llull
1st International Workshop on the Chinese Development Model
Huaxin Wang-Lu, Octasiano M. Valerio Mendoza IQS School of Management, Universitat Ramon Llull
Job Prospects and Internal Migration in China

Outline
1 Introduction
2 Research Gaps & Objectives
3 Theoretical Model
4 Data & Empirical Analysis
5 Main Results
6 Robustness Check
7 Concluding Remarks

A Brief Summary of Background
In the aspect of domestic migration:
The new Hukou reform (a household registration system) was
rolled out in late 2014 to allow more migrants permanently settle
down in small- and medium-sized Chinese cities.

Intra-provincial migration (across city boundaries) is evidenced as
being more important than inter-provincial migration (across
province boundaries) for both temporary and permanent
migration since around 2010 (Meng, 2020; Zhang & Zhao, 2013).

Intra-provincial migration (across city boundaries) is evidenced as
being more important than inter-provincial migration (across
province boundaries) for both temporary and permanent
migration since around 2010 (Meng, 2020; Zhang & Zhao, 2013).
Migrants were found consistently increasingly more educated
during 2000–2015 (only 2% migrants have at least a college degree in
1990, whereas 23.3% in 2015) (National Health and Family
Planning Commission, 2018).

In the aspect of industrial development:
The central government launched plans to promote nationwide
industrial upgrading, e.g., “Made in China 2025”.

The government also announced the “Vocational Education
Quality Improvement Action Plan” in 2020 to fill up skill gaps
(skilled technicians) that industrial transformation needs.

The government also announced the “Vocational Education
Quality Improvement Action Plan” in 2020 to fill up skill gaps
(skilled technicians) that industrial transformation needs.
A paradox in recent years that people cannot find jobs while
firms have difficulties to find workers (e.g., Athukorala and Wei,
2018; Chan, 2010) – the shut-down and relocation of numerous
factories as a result of rising labor costs (e.g., Chen et al., 2011; Lin,
2012; Wang et al., 2020); the jobless growth and the stagnation of the
creation of non-routine jobs threaten numerous types of work (Chen &
Xu, 2018; Frey & Osborne, 2017; Ge et al., 2021).

Research Gaps
Gap 1: Visionary migration studies are almost non-existent among
Chinese migration literature (partly due to data limitations).
Higher-educated migrants are more likely to make non-myopic
(looking forward, evaluate future economic conditions) migration
decisions; Liu and Shen (2014) suggest skilled migrants prioritize
career prospects over the quality of life in the decision-making
process.

Research Gaps
Gap 2:
The Hukou registration is primarily managed by prefecture cities
(sub-national administration unit below provinces), but migration
research on China’s internal migration is predominantly
inter-provincial – migrants moving across cities within the province of
origin were not counted.
And to our knowledge, even studies at the city level are either
cross-sectional or monadic (mainly due to data limitations).

Research Gaps
Gap 3:
Though several migration studies (Clark & Lisowski, 2017; Czaika,
2015; Yonemoto, 2021) borrow virtues, particularly, reference
dependence, from the prospect theory (Kahneman & Tversky, 1979),
the random utility maximization (RUM) model adopted by a body of
global migration literature hasn’t yet considered the
reference-dependent utility (at the very least not in a sequential
scenario).
However, previous literature indicates that changes in expectations
matter (e.g., Baumann et al., 2015; Shrestha, 2020).

Objectives & Approach
In sum:
to examine the effects of sector-based job prospects on
individual migration decisions across prefecture boundaries.

In sum:
(to overcome data limitations) created a unique quasi-panel of
66,427 individuals from 283 cities moving to 279 cities during
1997–2017.

In sum:
1997–2017.
(to overcome data limitations) constructed a proxy variable for job
prospects.

In sum:
1997–2017.
(to overcome data limitations) constructed a proxy variable for job
prospects.
introduced the reference-dependent utility to the sequential
RUM model developed by Bertoli et al. (2016).

The sequential model of migration (Bertoli et al., 2016) is as follows:
Uijk,t = wkt + βAkt+1(I) − cjk,t + ϵijk,t (1)
where wkt is the deterministic instantaneous component. β ∈ [0, 1) is the
time discount factor of the expected utility Akt+1(I). β = 0 represents
myopic decisions. I denotes that the location preference hinges on the
industry to which job categories that individual i searches for belong.
The reference-dependent migration value function (Czaika, 2015) is
as follows:
Vit = V k
it − V j
it = M(ỹk
it − ỹj
it) + N(yk
it+1, yj
it+1|yk
it , yj
it)
where N(·) = (yk
it+1 − yk
it − yj
it+1 + yj
it )α
= (∆it+1yk
− ∆it+1yj
)α
where M(·) is the reference-independent component. N(·) is the
reference-dependent utility where present economic situations in the origin
and destination city, i.e., yk
it and yj
it , respectively act as a reference point in
adjusting present expectations about the future, i.e., yk
it+1 and yj
it+1.

In a one-time migration scenario, our model on the foundation of
Czaika (2015) and Bertoli et al. (2016) can be expressed as:
Akt+g (I) =
1
βg
E[Akt(I)] =
1
βg
E[Akt(I)|rg
] · E[rg
] =
rg
βg
· ∆ityk
(I)
However, individuals can update their reference points after moving
to k at time t and choose any alternative location q among the choice
set D at time t + 1. If we assume that the stochastic component of utility
follows an i.i.d. Extreme Value Type-1 distribution (McFadden, 1974) with
zero mean where τ is the Euler constant, the recursive form in terms of
Equation (1) can be expressed as:
Uijk,t = wkt + β

τ + ln


X
q∈D
ewqt+1−ckq,t+1+β∆it+2yq
(I)



 − cjk,t + ϵijk,t

As McFadden (1974) shows, the probability of migrating from city j to
city k can be estimated as:
In

Pijk,t
Pijj,t

= wkt − wjt − cjk + β · [∆it+1yk
(I) − ∆it+1yj
(I)] (2)
where wjt is the utility for individual i choose to keep staying in city j at time
t.
We also made an extension by considering the log-linear function as
does Beine et al. (2019) to account for the non-linearity and constant
relative risk aversion discussed in Anderson (2011):
In

Pijk,t
Pijj,t

= wkt − wjt − cjk + β · ln
∆it+1yk
(I)
∆it+1yj (I)
(3)
In the empirical part, we estimated both (report the results of Eq. (3) here).

Outline Introduction Research Gaps Objectives Theoretical Model Data Empirical Analysis Main Results Robustness Check Concluding Remarks
Outline
1 Introduction
2 Research Gaps Objectives
3 Theoretical Model
4 Data Empirical Analysis
5 Main Results
6 Robustness Check
7 Concluding Remarks

Data Sources
the 2017 China Household Finance Survey (CHFS)

Data Sources
city-level longitudinal statistics (1996–2017) retrieved from the
China Data Institute

Data Sources
city-level longitudinal statistics (1996–2017) retrieved from the
China Data Institute
the China Hukou Registration Index (CHRI) developed by Zhang
et al. (2019)

Descriptive Statistics
Micro-level data:

Macro-level data:

Trending Indicator: accelerator / decelerator?
Trendingij,t =
Eij,t − Eij,t−1
Eij,t−1
−
Eij,t−1 − Eij,t−2
Eij,t−2
= ∆tGRij
Trendingik,t =
Eik,t − Eik,t−1
Eik,t−1
−
Eik,t−1 − Eik,t−2
Eik,t−2
= ∆tGRik
where the quantity of employment at time t in the sector where individual i
is employed is Eij,t for the origin city j and Eik,t for the destination city k.

Trending Indicator: accelerator / decelerator?
Trendingij,t =
Eij,t − Eij,t−1
Eij,t−1
−
Eij,t−1 − Eij,t−2
Eij,t−2
= ∆tGRij
Trendingik,t =
Eik,t − Eik,t−1
Eik,t−1
−
Eik,t−1 − Eik,t−2
Eik,t−2
= ∆tGRik
where the quantity of employment at time t in the sector where individual i
is employed is Eij,t for the origin city j and Eik,t for the destination city k.
Implications behind reference dependence:
Positive growth does not necessarily lead to better job prospects, and
negative growth over two years, such as GRij,t = −0.5% and
GRij,t−1 = −0.7%, ∆tGRij = 0.2% because the scale of the decline is
narrower, implying the job prospect stands a chance of getting better.

Empirical Analysis – Fixed Effects I
Baseline models with time, origin, destination and/or sector fixed effects to
address multilateral resistance to migration (in cross-sectional studies):
Mijk,t = α + β1Xit + β2DistanceT
(Job_Trendingijk,t ) + β3Distance(Zijk,t−1)
+γt + ϵijk,t
(Trendingijk,t ) + β3Distance(Zijk,t−1)
+γt + γj + γk + γs + ξijk,t
+γt + γjk + γs + ξijk,t

Empirical Analysis – Fixed Effects II
Models with time-location fixed effects to address multilateral resistance to
migration where the future attractiveness of alternative locations matters:
+ γjt + γk + γs + ξijk,t
+ γkt + γj + γs + ξijk,t
Based on the equations above, we further consider time-industry fixed
effects:
+γjt + γk + γst + ϵijk,t
+γkt + γj + γst + ϵijk,t

Robustness Check – Multilevel Logit GMM
log(
πit
1 − πit
) = α + β1Xit + β2DistanceT
(Trendingijk,t)
+β3Distance(Zijk,t−1) + γt + µC
where µC is assumed to be i.i.d normally distributed with a zero mean and
level-2 variance σC , accounting for the effects of being in city group C on the
log-odds that Mijk,t = 1. β1 is still level-1 unknow parameters and, β2 and β3
are level-2 parameters to be estimated. The index C can be either an origin
city or a destination city.
The two-step system GMM models are almost the same as the FE models but
only include time and sector fixed effects as the city-level fixed effects will be
automatically handled in the first differenced model.

Table: Estimation of Migratory Probabilities
OLS Multilateral Resistance
(1) (2) (3) (4) (5) (6) (7)
distance_trend 0.0876∗∗∗
0.1933∗∗∗
0.1408∗∗∗
0.1636∗∗∗
0.1491∗∗∗
0.1721∗∗∗
0.1321∗∗∗
(0.0297) (0.0290) (0.0330) ( (0.0579) (0.0329) (0.0576) (0.0321)
distance_CHRI 0.0222∗∗
(0.0087)
distance_trendXdistance_CHRI 0.2212∗
(0.1129)
pioneer 0.0667∗∗∗
0.0619∗∗∗
0.0576∗∗∗
0.0666∗∗∗
0.0596∗∗∗
0.0614∗∗∗
(0.0043) (0.0035) (0.0063) (0.0045) (0.0033) (0.0036)
Ind. Controls Y Y Y Y Y Y
City Controls Y Y Y Y Y Y
Constant 0.0090∗∗∗
0.0622∗∗∗
0.0603∗∗∗
0.0406∗∗∗
0.0396∗∗∗
0.0607∗∗∗
0.0600∗∗∗
(0.0009) (0.0045) (0.0041) (0.0038) (0.0042) (0.0042) (0.0042)
Time FE Y Y Y Y Y
Industry FE Y Y Y Y Y
Origin FE Y Y Y
Destination FE Y Y Y
Pairs of cities FE Y
Origin-year FE Y
Dest-year FE Y
Obs 969998 749219 729965 408877 729960 729769 729725
Notes: Standard errors shown in parentheses are clustered at the destination city. ∗
p 0.10, ∗∗ p 0.05, ∗∗∗ p 0.01

Table: Multilevel Logit and Two-step System GMM
Two Level Logit Three Level Logit GMM
(1) (2) (3) (4) (5) (6) (7) (8) (9)
distance_trend 5.3903∗∗∗
4.5964∗∗∗
3.3363∗∗∗
5.3922∗∗∗
4.5847∗∗∗
3.3363∗∗∗
0.3157∗∗∗
0.2493∗∗∗
0.2751∗∗∗
(1.0279) (0.6027) (1.0040) (1.0181) (0.6120) (1.0040) (0.0754) (0.0569) (0.0582)
Ind. Controls Y Y Y Y Y Y Y Y Y
City Controls Y Y Y Y Y Y Y Y Y
Intercept -1.839∗∗∗
-1.5841∗∗∗
0.0077 -1.9504∗∗∗
-1.7681∗∗∗
0.0077 0.0610∗∗∗
0.0590∗∗∗
0.0590∗∗∗
(0.3101) (0.2573) (0.2468) (0.3137) (0.2857) (0.2468) (0.0046) (0.0046) (0.0046)
Level 2 var. 0.1484 1.6611 1.2684 0.1869 0.2416 1.99e-33
(0.0372) (0.1671) (0.1009) (0.0312) (0.0455) (3.38e-34)
Level 3 var. 0.1054 1.6046 1.2685
(0.0398) (0.1685) (0.1009)
ICC 0.0432 0.3355 0.2783 0.0816 0.3595 0.2783
(0.0104) (0.0224) (0.0160) (0.0112) (0.0214) (0.0160)
Nest origin destination pair origin destination pair
sub-Nest education education education
Time FE Y Y Y Y Y Y Y Y Y
Industry FE Y Y Y
AR(2) 0.724 0.790 0.790
Hansen’s J test 0.349 0.263 0.404
Obs 749219 749219 749219 749219 749219 749219 729965 729965 729965
Notes: Standard errors shown in parentheses are clustered at the destination city. ∗ p
0.10, ∗∗ p 0.05, ∗∗∗ p 0.01

Graphs
0.0018 0.0045 0.0110
0.0267
0.0626
0.1378
0.2745
0
.1
.2
.3
.4
.5
Effects
on
Marginal
Predicted
Mean
-.7 -.5 -.3 -.1 .1 .3 .5
distance_trend
Conditional Marginal Effects with 95% CIs
(a) Two-level random intercept model.
0.0019 0.0046 0.0114
0.0274
0.0636
0.1378
0.2704
0
.1
.2
.3
.4
.5
Effects
on
Marginal
Predicted
Mean
-.7 -.5 -.3 -.1 .1 .3 .5
distance_trend
Conditional Marginal Effects with 95% CIs
(b) Three-level random intercept
model.
Figure: Marginal effects of destination-nested model.
Notes: All results are statistically significant at the 1% level.

Conclusions
Macro surroundings often silently yet profoundly influence
individual perceptions. This paper mirrors the role of contextual
evolution in forming expectations of all relevant individuals.

Conclusions
Most primarily, we show that a 10% increase in the ratio of
sector-based job prospects in cities of destination to cities of
origin raises the probability of migration by 1.281–2.185
percentage points, and the effects tend to be stronger when the
scale of the ratio is larger.

Conclusions
Having a family migration network causes an increase of
approximately 6 percentage points in migratory probabilities.

Conclusions
Having a family migration network causes an increase of
approximately 6 percentage points in migratory probabilities.
Labor migrants are more likely to be male, unmarried, younger,
or more educated.

Contribution
The contribution is fourfold:
Theoretically, we expanded the RUM model of migration by
synthesizing the virtues of dynamic discrete choice models and
reference-dependence.

Contribution
The contribution is fourfold:
Theoretically, we expanded the RUM model of migration by
synthesizing the virtues of dynamic discrete choice models and
reference-dependence.
Empirically, 1. we created a unique quasi-panel to account for
city-level dyadic longitudinal effects alongside controlling for
important individual (and household) characteristics,
2. we addressed multilateral resistance to migration that both
cross-sectionally and longitudinally, and
3. we explored individual migration decisions with generic job
prospects in relation to regional industrialization.

Thanks Ending

Anderson, J. E. (2011). The Gravity Model. Annual Review of Economics,
3(1), 133–160.
Athukorala, P.-c., Wei, Z. (2018). Economic Transition and Labour
Market Dynamics in China: An Interpretative Survey of the
‘Turning Point’ Debate. Journal of Economic Surveys, 32(2),
420–439.
Baumann, R., Svec, J., Sanzari, F. (2015). The Relationship between
Net Migration and Unemployment: The Role of Expectations.
Eastern Economic Journal, 41(3), 443–458.
Beine, M., Bourgeon, P., Bricongne, J.-C. (2019). Aggregate
Fluctuations and International Migration. The Scandinavian
Journal of Economics, 121(1), 117–152.
Bertoli, S., Brücker, H., Fernández-Huertas Moraga, J. (2016). The
European crisis and migration to Germany. Regional Science
and Urban Economics, 60, 61–72.
Chan, K. W. (2010). A China Paradox: Migrant Labor Shortage amidst
Rural Labor Supply Abundance. Eurasian Geography and
Economics, 51(4), 513–530.

Chen, S., Jefferson, G. H., Zhang, J. (2011). Structural change,
productivity growth and industrial transformation in China.
China Economic Review, 22(1), 133–150.
Chen, Y., Xu, D. (2018). The Impact of the ’Artificial Intelligence
Revolution’ on Employment: Review and Prediction. SSRN
Electronic Journal.
Clark, W. A. V., Lisowski, W. (2017). Prospect theory and the
decision to move or stay. Proceedings of the National Academy
of Sciences, 114(36), E7432–E7440.
Czaika, M. (2015). Migration and Economic Prospects. Journal of
Ethnic and Migration Studies, 41(1), 58–82.
Frey, C. B., Osborne, M. A. (2017). The future of employment: How
susceptible are jobs to computerisation? Technological
Forecasting and Social Change, 114, 254–280.
Ge, P., Sun, W., Zhao, Z. (2021). Employment structure in China
from 1990 to 2015. Journal of Economic Behavior
Organization, 185, 168–190.

Kahneman, D., Tversky, A. (1979). Prospect theory: An analysis of
decision under risk. Econometrika, 47, 263-291..
Econometrica, 47(2), 263–292.
Lin, J. Y. (2012). From Flying Geese To Leading Dragons: New
Opportunities and Strategies for Structural Transformation
in Developing Countries. Global Policy, 3(4), 397–409.
Liu, Y., Shen, J. (2014). Spatial patterns and determinants of skilled
internal migration in China, 2000-2005. Papers in Regional
Science, 93(4), 749–771.
McFadden, D. (1974). Conditional Logit Analysis of Qualitative Choice
Behaviour. In P. Zarembka (Ed.), Frontiers in econometrics
(pp. 105–142). Academic Press New York.
Meng, L. (2020). Permanent migration desire of Chinese rural
residents: Evidence from field surveys, 2006–2015. China
Economic Review, 61(January), 1–15.
National Health and Family Planning Commission. (2018). Report on
China’s Migrant Population Development, 2018 (tech. rep.).
China Population Publishing House.

Shrestha, M. (2020). Get Rich or Die Tryin’: Perceived Earnings,
Perceived Mortality Rates, and Migration Decisions of
Potential Work Migrants from Nepal. The World Bank
Economic Review, 34(1), 1–27.
Wang, F., Xia, J., Xu, J. (2020). To upgrade or to relocate? Explaining
heterogeneous responses of Chinese light manufacturing
firms to rising labor costs. China Economic Review, 60, 101333.
Yonemoto, K. (2021). Reference-dependent preference and
interregional migration: extending the Harris–Todaro model.
Letters in Spatial and Resource Sciences, 14(1), 1–10.
Zhang, J., Wang, R., Lu, C. (2019). A quantitative analysis of Hukou
reform in Chinese cities: 2000-2016. Growth and Change, 50(1),
201–221.
Zhang, J., Zhao, Z. (2013). Measuring the income-distance tradeoff
for rural-urban migrants in China. IZA Discussion Paper,
(7160).

Job Prospects and Internal Migration in China

Recommended

Recommended

More Related Content

Similar to Job Prospects and Internal Migration in China

Similar to Job Prospects and Internal Migration in China (20)

Recently uploaded

Recently uploaded (20)

Job Prospects and Internal Migration in China