The existence and persistence of household financial hardship: A Bayesian multivariate dynamic logit framework
Thoughts
Head empty
Annotations
brown2014 - p. 285
To be specific, the existing literature on household finances has generally focused on financial decision-making in the context of the nature and characteristics of the financial portfolios held including decisions regarding stock market participation and the diversification of financial assets (see Campbell, 2006).
brown2014 - p. 285
Our analysis of financial problems thus sheds light on an area of household finances, which has attracted surprisingly little attention in the existing literature.
brown2014 - p. 286
The econometric framework is described below in four steps. The first step relates to the specification of the incidence of the kth financial problem of the ith household at time t within a joint modelling framework. The second step concerns modelling the interdependence of the incidence of the different financial problems and how these interact with each other since the overall financial hardship of a household is a combination of each of these effects. We do this in two ways: firstly, by allowing for the dynamic aspect of the incidence of each financial problem; and, secondly, by explicitly modelling unobserved household heterogeneity, allowing for correlation between the different financial problems. The third step involves modelling missing observations using a logit model. The final step entails the construction of the joint likelihood of the financial problems of all households in the sample bringing together all three extensions outlined above.
brown2014 - p. 288
We use the British Household Panel Survey (BHPS), a survey conducted by the Institute for Social and Economic Research comprising approximately 10,000 annual individual interviews. For wave one, interviews were conducted during the autumn of 1991.
brown2014 - p. 288
We focus on household reference persons as they are responsible for meeting primary household expenditures, such as housing payments, and play the primary role in household financial decision making (we refer to them as household heads for brevity).6
brown2014 - p. 288
Firstly, information is available in all waves relating to whether households over the last 12 months have had any difficulties paying for their accommodation (denoted fprob1). Secondly, information was gathered on the extent to which households experienced financial problems relating to loans (denoted fprob2). Thirdly, in the BHPS from 1996 onwards, information on financial hardship at the household level can be discerned from the responses of the head of household regarding the ability of the household to: afford to keep their home adequately warm (denoted fprob3); be able to pay for a week’s annual holiday (denoted fprob4); replace worn-out furniture (denoted fprob5); be able to buy new, rather than second-hand, clothes (denoted fprob6); be able to eat meat, chicken, fish every second day (denoted fprob7); and be able to have friends or family for a drink or meal at least once a month (denoted fprob8). Finally, information is available indicating whether the household is unable to save anything on a monthly basis (denoted by nosave).
brown2014 - p. 289
However, there is some evidence that financial hardship was starting to increase in 2008, which coincides with the start of the recent global financial crisis.
Note: Important interface (propagation effect?) between microeconomic and macroeconomic financial stress
brown2014 - p. 290
Furthermore, there is potential interdependence between the different types of financial hardship experienced by the household. For example, experiencing a particular type of financial problem in the past may lead to the household experiencing a different type of financial problem in the current period.
brown2014 - p. 291
In terms of overall model performance, the calculated log Bayes factor is 34.02, giving very strong support for rejecting the null hypothesis that the slope parameters are jointly equal to zero, see Kass and Raftery (1995).
brown2014 - p. 291
In Table 4 Panels A to C, we present the results from estimating the system of nine logit equations of financial hardship.
brown2014 - p. 291
Panel A presents the estimates associated with the dynamic process of the dependent variables. Persistence in financial problems, as indicated by a statistically significant positive estimated effect on the relevant lagged dependent variable, is found for experiencing problems paying for accommodation, problems with loan repayments, affordability issues with annual holidays, new furniture and entertaining family and friends as well as being unable to save on a monthly basis. With the exception of entertaining friends and family, it is apparent that the financial problems characterised by the most persistence are those associated with the types of expenditure that are often financed by credit such as loans, mortgages and credit cards. In contrast, the categories of financial problems characterised by the least persistence are those associated with expenditure on food, clothes and heating, which are generally paid for with cash/debit cards rather than via the use of credit. The largest effect is found for problems with loan repayments, where, if the same problem was experienced in the previous year, the likelihood of it occurring in the current period increases considerably. Hence, the relative probability of currently reporting problems with loan repayments, conditional on whether they were experienced in the previous year, is 78%.