commuting patterns Micro-evidence for the Netherlands Stefan Groot - PowerPoint PPT Presentation

The educational bias in commuting patterns Micro-evidence for the Netherlands Stefan Groot VU University Amsterdam Co-authors: Henri de Groot Paolo Veneri The Hague, March 13, 2013

Contents of the presentation  Research questions  Theory  Data  Stylized facts  Selected empirical results  Conclusions

Research questions  Analyze the relation between education and commuting behavior of Dutch workers  Attempt to separate the effects of education from the effects of income  Try (to some extent) to explain observed commuting patterns

There could be a bias in commuting patterns across several dimensions:  Housing location  Work location  Commuting distance  Mode of transport  Commuting can be considered as the outcome of an optimization problem

Previous literature  Commuting time paradox ◦ Van Ommeren and Rietveld (2005)  Individual attributes account for a large part of commuting behavior ◦ Giuliano and Small, 2993  Education is associated to longer commutes ◦ Shen, 2000; Lee and McDonald, 2003; Papanikolaou, 2006  Higher educated more likely to be long distance commuters ◦ Öhman and Lindgren, 2003

Why do we expect commuting to be related to level of education?  Demand and supply on the housing market ◦ Higher willingness to pay (interaction with income) ◦ Higher educated may prefer suburbs ◦ Revival of cities (Glaeser and Saiz, 2004; Glaeser, 2011) ◦ Higher educated are more likely to be house owner  Van Ommeren and Leuvensteijn (2005): 1 percent-point increase in transaction costs decreases mobility by 8 percent ◦ Relatively high supply of social housing in large cities

Why do we expect commuting to be related to level of education?  Labor market search frictions ◦ Excess commuting (Van Ommeren and Van der Straaten, 2008) ◦ Work of higher educated is more specialized ◦ Higher search frictions  Agglomeration economies  Possibility for leisure / work during commute

Data  Micro data from Statistics Netherlands (CBS)  SSB Banen + labor force survey (EBB)  Apply selection criteria (wage, fte, age)  Source of residence location is always the GBA register (available for all Dutch residents).  Source of work location is EBB  Spatial level: municipality

Combine SSB and EBB for commuting matrix  Use register data to obtain total employment and total working residents by municipality.  Use labor force survey to fill commuting matrix  Apply RAS method to guarantee consistency between commuting matrix and row and column totals.

On the maps that will appear in a few seconds:  Higher educated workers: those with higher tertiary education (HBO / University degree)  Lower educated workers: the rest  Balance index = (inflow-outflow) / (inflow+outflow)  Only largest commuter flow between two municipalities is presented

Balance index and commuter flows of higher educated workers

Balance index and commuter flows of lower educated workers

Relation between land rents and balance index Left: highly educated, right: lower educated 800 800 700 700 Amsterdam Amsterdam Land rent (euro/m2) Land rent (euro/m2) The Hague 600 The Hague 600 500 500 Utrecht Utrecht 400 400 Rotterdam Almere Rotterdam 300 300 Almere Zaanstad Zaanstad Eindhoven Eindhoven Groningen Groningen Breda Breda Tilburg Tilburg 200 200 Nijmegen Nijmegen Enschede Enschede Apeldoorn 100 Apeldoorn 100 0 0 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 Balance index of highly educated workers Balance index of lower educated workers

Regression Dependent variable: balance index Lower educated Higher educated All workers workers workers N (observations) 437 437 437 Log population 0.147 *** 0.199 *** 0.165 *** Log population density – 0.015 0.060 ** 0.015 Wage residual 0.096 0.596 * 0.451 Land rent 0.090 *** – 0.094 *** 0.012 R-squared 0.272 0.362 0.333

Stylized facts by level of education Type of education Private transport Public transport %-share distance time %-share distance time Primary education 10.5 15.5 15.7 38.0 92.0 8.0 Lower secondary education (VMBO, MBO 1) 12.3 16.6 21.6 41.1 92.7 7.3 Higher secondary education (HAVO, VWO) 15.3 20.4 27.0 45.2 87.2 12.8 Lower tertiary education (MBO 2, 3) 13.5 17.2 24.3 41.8 93.0 7.0 Lower tertiary education (MBO 4) 15.5 19.2 28.6 45.4 93.7 6.3 Higher tertiary education (HBO, BA) 17.5 21.8 33.4 49.5 91.3 8.7 Higher tertiary education (MA, PhD) 20.0 24.9 41.1 54.3 82.6 17.4

Regressions individual commuting behavior  Dependent: individual commuting time, distance, or mode of transport  Methodology o OLS for time / distance o Multinomial logit for mode of transport  Independents: characteristics of individual, job, work and residence location  Robust when including work and residence location fixed effects

Selected empirical results: distance / time  Females and older workers commute less  Higher incomes commute further and faster  Education explains more than wage  Higher educated workers have ceteris paribus longer commutes o Particularly university graduates  Lower educated commute further when earning higher wages, higher educated commute far anyway

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 education dummies (right) on log distance Interaction of education and log wage (left) and Primary education Lower secondary education (VMBO, MBO1) Lower tertiary education (MBO2+3) Lower tertiary education (MBO4) Higher secondary education (HAVO, VWO) Higher tertiary education (HBO, BA) Higher tertiary education (MA, PhD) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Primary education Lower secondary education (VMBO, MBO1) Lower tertiary education (MBO2+3) Lower tertiary education (MBO4) Higher secondary education (HAVO, VWO) Higher tertiary education (HBO, BA) Higher tertiary education (MA, PhD)

Selected empirical results: distance / time  Workers commuting to jobs in densely populated areas commute only slightly further, but much slower  Residents of densely populated areas commute less  Workers commuting towards more productive areas commute much further  Residents of expensive locations commute less

Selected empirical results: mode of transport  Females are more likely to commute by car, less by bike  High wage earners use more cars, less bikes  Apart from the obvious (walking, cycling), distance has a particularly strong impact on the use of trains  No relation between sector and use of public transport  Higher educated workers are ceteris paribus more likely to use trains or cycle, less likely to commute by car

Conclusion  Substantial heterogeneity in commuting patterns  Higher educated workers commute further and longer  Higher educated workers are ceteris paribus more likely to use bycicle/trains, less likely to commute by car  Effect of education goes beyond wage or commuting distance  Higher educated workers are ceteris paribus more likely to commute from regions with high amenities

Questions?