RDP 2014-13: Mortgage-related Financial Difficulties: Evidence from Australian Micro-level Data Appendix E: Household-level Data

In the 2006 and 2010 HILDA Survey's wealth modules, respondents were asked, ‘Do you currently own or have you ever owned a residential property?’ If a respondent answered yes, they were then asked, ‘During the last 12 months have you (or your household) been unable to meet a mortgage payment by the due date on any housing or property loan because of financial difficulties?’ Respondents could answer ‘yes’, ‘no’, ‘did not own/have a property loan during last 12 months’ or ‘don't know’. Thus, only respondents that had a property loan in the previous 12 months should have responded ‘yes’ or ‘no’ to the second question. However, around half of households that did not have any property loans (including mortgages for non-owner-occupied property) at the time of the survey responded ‘no’. Although some of these respondents may have had a property loan during the previous 12 months but repaid it before the interview date, the numbers appear too large to solely reflect this. As a consequence, the share of households that missed a mortgage payment could be understated. To address this problem, in our descriptive analysis we restrict the sample to households that had a mortgage on their residence at the time of the survey or at the time of the previous survey.

This issue should not affect the dependent variable in our probit model, as the inclusion of explanatory variables related to the DSR and LVR implicitly restricts the estimation sample to those households that had owner-occupier housing debt in 2009. However, this issue makes it necessary to recode the missed payments variable from the 2006 survey before including it as a categorical explanatory variable. Households that responded ‘no’ to missing a mortgage payment in 2006 but did not appear to have a property loan in the 2005 or 2006 surveys are treated as non-responders. The non-response category also includes households that: refused to answer the question; reported not knowing the answer to the question; were part of the survey sample in 2006 but did not respond; or were not part of the survey sample in 2006. Recoding the missed payments variable in this way ensures that the estimation sample is not unnecessarily restricted to households that responded to both 2006 and 2010 surveys. It also avoids the measurement error that would occur when classifying households that did not have a property loan in 2006 as having not missed a payment.