Abstract
Using unique data collected in October–December 2012 we estimate the link between commuting for work and level of individual exposure to floods. We find that commuters on average have higher earnings than non-commuters. Individuals affected by one flood commute by 11.2% more than unaffected individuals. We conjecture that the increase is linked to intentions to cover flood-related losses, decrease households' vulnerability to flood risk or out-migrate from the risk areas. Individuals affected by at least two floods are by 20.2% less likely to commute relative to those unaffected. We explain this non-linear effect by the fact that many households out-migrate after the first flood. Stayers commute less, because they are different from non-stayers in some underlying characteristics related to education, employment and family circumstances, which strongly affect commuting behaviour. We further find that in a commuting family an individual is by 53.8% more likely to commence commuting relative to a non-commuting family. Choice of commuting destination is often similar to that of other family members.
Acknowledgements
We are grateful to Oliver Bakewell, Jeanette Schade and Thomas Faist for their comments during the conference at Bielefeld University. We are also thankful to reviewers for their recommendations.
The financial support of grants CZ.1.05/1.1.00/02.0073, CZ.1.07/2.4.00/31.0056 and LD 13032 is kindly acknowledged.
Notes
1 In a related study, Brázdil et al. (Citation2011) research the Morava river, however the Becva river remained rather unexplored in their analysis.
2 A typical insurance contract has separate provisions for insuring a house (walls, doors, cellar, etc.) and assets in the house (boiler, furniture, electronics, etc.).
3 An alternative procedure would be to construct a likelihood function in the spirit of the Tobit model.
4 This instrument indeed has a significant predictive power. If family size increases by one individual, the probability of reporting income drops by 5.2%. This estimate is significant at 5% significance level.