Commuting and Minimum wages in Decentralized Era Case Study from Java Island Raden M Purnagunawan
Outline 1. Introduction 2. Brief Literature review 3. Data Source and Construction 4. The aggregate commuting in Java 5. Individual Economic Returns to Commuting 6. Conclusion
Introduction Minimum wages variation across region is likely to attract labor movement to region that offer higher wages The variation of minimum wages between districts in Indonesia (in this case Java) is quite large, either using nominal or real minimum wage adjusted by the consumer price index How the differences in minimum wages across districts, distance and other regional characteristics affect the commuting pattern between districts at the aggregate level? How this mobility of labour affect individual wages?
District-level Monthly Minimum Wages in Java, 2001 2007 (Rp 000) 2001 2002 2003 2004 2005 2006 2007 Nominal Minimum Wages Average (mean) 283.6 368.6 412.6 446.2 477.2 561.6 623.3 Std. Deviation 65.0 102.3 104.0 109.0 113.0 126.6 140.3 Minimum 220.0 245.0 274.0 310.0 336.0 390.0 448.5 Maximum 426.3 591.3 635.0 673.0 713.0 835.9 905.0 Range 206.3 346.3 361.0 363.0 377.0 445.9 456.5 Coefficient of Variation 22.9 27.8 25.2 24.4 23.7 22.5 22.5 Real Minimum wages (2002 = 100) a Average (Mean) 298.2 351.8 377.8 385.6 353.6 388.9 405.0 % increase 18.0 7.4 2.1-8.3 10.0 4.1 Std. Deviation 68.6 98.3 93.9 92.9 81.3 86.1 89.9 Minimum 230.0 234.1 252.2 268.5 251.2 273.0 294.5 Maximum 448.6 570.5 576.8 579.3 529.1 574.2 595.3 Range 218.7 336.4 324.6 310.7 277.9 301.2 300.8 Coefficient of Variation 23.0 28.0 24.9 24.1 23.0 22.1 22.2
Research Questions Does commuting vary substantially and systematically by location and by personal characteristics? To what extent can the minimum wages explain the aggregate commuting pattern between districts and the decision to commute at the individual level? Is there any wage differential between commuters and non-commuters?
Brief Literature Review Most of the literature on labour mobility focuses on permanent migration, long or short term, but deals less with temporary migration, especially commuting and circular migration due to data availability Determinants of commuting can also be inferred from the studies of migration, as commuting can be seen as a special case of labour mobility (Eliasson et al. 2003) The main determinants of migration (Greenwood 1975, 1997): distance, income, the psychic cost of migration, information, the unemployment rate and personal characteristics.
Brief Literature Review Empirical studies Hazan (2004) for 3 Baltic Countries on the relationship between commuting and income disparities Commuting narrowed the wage gap between capital cities and rural, especially in Estonia and Latvia but only in some parts of Lithuania Commuter gain 40 93 percent wage premium
Brief Literature Review Indonesian Cases: Hugo (1977) collected data from 14 villages in West Java in 1973 Most of the circular migrant work on labour intensive, low productivity small scale informal economy Earn meager level of urban earning, but could remit 21 to 44 percent of their incomes Hetler (1989) in-depth data from 1 village in Central Java Importance of circular migrant in increasing the family income in the village, especially through remittances. Village economy has benefited from urban informal sector activities. Remittances from circular migration have significantly increase the household income and wellbeing.
Data Source and Construction Main Source : Indonesian Labour Force Survey (Sakernas) 2007 Sample sizes covers 285,989 households and 793.380 individuals aged 10 and above Sample for Analysis Male aged 15 and above Those who work and reside in Java Has district level minimum wage High population density with relatively better infrastructure and lower transportation cost Remove missing data Final Sample: 61,630 male worker. (9,382 men or 15.2% commuter)
Data Characteristics and Limitation Very large sample Commuting data only available for 2007 data Commuting data only for those who have job Has information on commuting frequency No information on commuting time and distance No information on specific location of workplace (urban/rural) No information on whether an in individual live near border area No household unique ID (should be available)
Commuter Commuters are defined as employed persons whose workplace is located in a district other than their place of residence, Workers who reside and work in the same district are defined as non-commuters
The Aggregate Commuting
Net commuting in Java, 2007
Estimation Strategy Multiplicative form of modified gravity-type model as suggested by (Silva and Tenreyro, 2006) estimated using a pseudo-maximum-likelihood (PML) Naturally deals with zero values of the dependent variable N N Mij = exp θ0 + θ1 ln Pi + θ2 ln Pj + θ3 ln Dij + ϕ n ln Xin + δ n ln X jn εij n= 1 n= 1
Variable description and summary statistics Variables description Mean Std. Dev. Min Max Total Commuters (M) 474.8 1369.8 0 29509 Commuters in the formal Sector 205.8 594.3 0 11817 Commuters in the informal Sector 269.0 921.5 0 17692 Minimum wages (MW) (Rp. 000) 659.9 157.1 448.5 905.0 Working age population (POP) (000) 987.8 583.4 12.8 2,947.8 Distance (D) 129.0 103.2 5.4 525.3 Current annual GDP growth 2007 (GDPGROWTH) 5.5 1.3 0.8 13.7 Unemployment rate (UNEMPLOY) 10.4 4.2 2.0 22.2 Proportion of urban population (URBAN) 59.6 33.2 4.9 100 Dummy 1 for shared border, 0 otherwise (CONTIG) 0.05 0.21 0 1 Source: Author s calculation from NLS 2007 (August round) using weights. Observations only include a pair of districts that have commuting activity, eitherinward, outward or both.
Poisson Pseudo Maximum Likelihood result of the modified gravity model of commuting in Java, 2007 All Formal Informal log_pop i 0.731*** 0.507*** 0.963*** log_pop j 0.353*** 0.274*** 0.447*** log_mw i -1.984*** -0.781** -3.109*** log_mw j 3.062*** 2.953*** 3.077*** log_distance ij -0.394*** -0.519*** -0.272*** log_gdpgrowth i -0.264-0.080-0.385** log_ GDPGROWTH j 0.200 0.219 0.208 log_unemploy i 0.411** 0.118 0.642*** log_ UNEMPLOY j -0.544*** -0.470** -0.620** CONTIG 0.069-0.110 0.246 log_urban i -0.342*** -0.304*** -0.383*** log_ URBAN j 1.079*** 1.216*** 0.973*** N 2,304 2,304 2,304 Wald chi 2 795.263 468.734 677.024 Pseudo R 2 0.524 0.417 0.482
On the results Distance acts as a significant deterrent to commuting The deterrent effect of distance is higher for the formal workers than for informal workers Suggest informal workers are more willing to travel longer distances and consequently pay higher travel costs than formal workers (assuming that distance reflects the cost of travelling) Higher unemployment in the destination region was a strong deterrent to commuting for workers in both the formal and informal sectors, while a higher unemployment rate in the place of origin induced workers to commute. Minimum wage is strongly correlated with gross commuting in Java, where a district s minimum wages act as a strong pull-factor
Individual Economic Returns to Commuting
Table 3 Distribution of commuter and non-commuter workers by age groups, education attainment and working status in Java, 2007 Urban Rural Total Non Commuter Commuter Total Non Commuter Commuter Total Non Commuter Commuter Total Number of workers a 11,503 2,418 13,921 9,719 1,518 11,237 21,222 3,937 25,158 Distribution (%) 82.63 17.37 100.00 86.49 13.51 100.00 84.35 15.65 100.00 By age group 15 24 15.17 12.49 14.70 15.95 14.61 15.77 15.53 13.31 15.18 25 34 29.39 30.50 29.58 27.08 30.18 27.50 28.33 30.38 28.65 35 49 37.99 43.77 38.99 37.24 42.44 37.94 37.64 43.26 38.52 50 64 14.94 12.47 14.51 16.35 12.19 15.79 15.59 12.36 15.08 65 + 2.51 0.77 2.21 3.39 0.58 3.01 2.91 0.70 2.57 100 100 100 100 100 100 100 100 100 By education attainment Less than Primary 8.22 2.99 7.32 17.85 10.36 16.84 12.63 5.83 11.57 Primary 30.63 21.27 29.00 48.69 48.33 48.64 38.90 31.71 37.78 Junior Secondary 21.78 15.54 20.69 18.78 22.42 19.27 20.40 18.20 20.06 Senior Secondary 30.05 39.13 31.63 11.10 15.55 11.70 21.37 30.03 22.73 Tertiary 9.32 21.06 11.36 3.57 3.35 3.54 6.69 14.23 7.87 100 100 100 100 100 100 100 100 100 By working status Formal 53.33 76.65 57.38 30.92 43.55 32.62 43.07 63.88 46.32 Informal 46.67 23.35 42.62 69.08 56.45 67.38 56.93 36.12 53.68
Empirical Model for Individual Economic Returns to Commuting Based on standard Mincerian Earning Function Allow for Endogeneity in Commuting decision Log Y ij = α + β X ij + δ COMMUTE ij + ε ij (1) COMMUTE ij = θ + ϕmwdiff i,rw + σz ij +µ ij (2) X ij is a vector of personal, occupational, sectoral, demographic and regional characteristics, Work status (j): formal or informal worker
Table 3. Variables definitions and sample means Variable Definition Whole sample Commuter Non Commuter Difference INCOME Monthly Income (in thousand Rupiahs) 890.67 1,382.6 807.3 575.3*** EDUC3 Dummy = 1 for those who has primary 0.37 0.29 0.38-0.092*** school education EDUC4 Dummy = 1 for those who has junior 0.20 0.18 0.20-0.027*** secondary school education EDUC5 Dummy = 1 for those who has higher 0.23 0.32 0.22 0.099*** secondary school education EDUC6 Dummy = 1 for those who has university graduate education 0.08 0.16 0.07 0.086*** AGE Age 38.42 37.31 38.61-1.295*** EXPER Experience (Age-Years of schooling -6 ) 24.03 21.54 24.45-2.905*** EXPER2 EXPER2/100 7.63 5.96 7.92-1.955*** Number of Observation 61,630 9,382 52,248
The Instrument MWDiff difference between the highest value of minimum wages in districts adjacent to the district of residence with the minimum wage in the district where the individual resides can be seen as the potential wage premium or incentive for commuting, which is expected to affect the decision to commute could only affect individual hourly income through the decision to commute, as only those who commute could benefit from the incentive or wage premium in the minimum wage differential
Income effect of commuting OLS Instrumental Variable (IV) Formal Informal Formal Informal [1] [2] [3] [4] COMMUTE 0.187*** 0.163*** 0.753*** 0.481*** (0.008) (0.012) (0.151) (0.157) N 29,386 32,244 29,386 32,244 r2 0.461 0.280 0.376 0.266 Summary of first stage regression Excluded Instrument MWDIFF 0.004*** 0.003*** (0.000) (0.000) F Statistics (excluded instrument) 140.206 208.044 Endogeneity test of endogenous regressors 16.131*** 4.237***
On the results The coefficient of COMMUTE from the instrumental variable method is significantly higher than for the results in the OLS estimate It is likely that workers who based their decision to commute on MWDiff are low income rather than high income workers (LATE) likely to experience a higher wage premium than the average worker. might also overestimate the returns to commuting, as commuting is also a choice for higher income workers that choose to live in districts with a lower cost of living (which is also reflected by the minimum wage)
Conclusion Most of the cities in Java experienced a net inflow of commuters (especially in the metropolitan city of Jakarta), while the labour market in regencies experienced a net outflow. The agregate results suggest that distance and a high unemployment rate in district of destination acts as deterrent to commuting, as suggested in the migration literature. The results also suggest the magnitude of commuting from and into a district was strongly affected by minimum wages and urbanisation rate. Commuters enjoyed a significantly higher income than similar non-commuters. Compared to similar non-commuters, commuters in the formal sector make a higher percentage gain in the wage premium than those working in the informal sector Commuter from urban area had higher percentage of wage premium than those from rural area.
Limitation and Extension Data limitation The income premium might be overestimated we did not controlled for the cost of commuting Cannot differentiate the specific location of the workplace and the specific travel time and distance, Only includes Java to include permanent migration in the analysis what is the relationship between migration and commuting how do minimum wages and other regional characteristics affect the decision to commute or migrate These issues are left for the future research
Thank You