This paper studies thechannels through which limited attention and brand familiarity affect consumer behaviour. Using unique combination of datasets on UK mortgage transactions, advertising, product data and pre-existing links between lenders and borrowers through other financial products,I develop a structural discrete choice model with limited attention to identify which characteristics of suppliers and products affect (a) the likelihood of the alternative being considered and (b) preferences for the alternatives that are considered.
I identify two distinct types of borrowers. Type 1, which has characteristics commonly associated with lower financial sophistication (e.g. lower income or worse credit history), tends to be less price-sensitive and more inattentive. It is much more likely to consider lenders with which they have a pre-existing checking account than ’unfamiliar’ lenders. The wealthier and more educated Type 2 also displays inattention but a lesser extent and their consideration of alternatives is much less influenced by existing relationships with lenders. Both types show a strong preference for familiar lenders when choosing amongst the options they consider. They are willing to trade off nearly 5% of their post-tax annual income for taking out a mortgage with a lender with which they already have a checking account. In a counterfactual simulation I show that due to this strong ’brand loyalty’, a hypothetical intervention to make borrowers aware of all available alternatives has only a surprisingly modest effect on market outcomes.

Against a backdrop of rising household leverage during a period of falling interest rates, concerns have been raised about the risk of borrowers falling into financial distress. Although a number of studies have investigated how the composition of consumer debt varies across individuals and the experiences of those in financial distress, limitations with the survey datasets used have made it hard to understand these patterns in detail.
We exploit a rich administrative dataset containing the credit files of a large, representative sample of UK borrowers to investigate this topic. We use statistical cluster analysis to identify four data-driven subgroups of borrowers: mortgage-holders, standard-cost borrowers, high-cost borrowers and those with household bills only. We then analyse for each cluster the incidence of distress, how this links to personal characteristics and credit usage six months prior to distress, and some features of the path from there into difficulty.
We find incidences of distress varies markedly across clusters, with high-cost borrowers more than twice as likely to get into difficulty. Those who go on to experience distress tend to share some common characteristics six months prior to hitting problems, regardless of cluster: they are typically younger, lower income and have higher debt. But we also find interesting differences by cluster. For example, for mortgage-holders, those that fall into distress actually have slightly lower total debt balances than the nondistressed; for high-cost borrowers, income is actually slightly higher for the distressed than the non-distressed. There are also commonalities across cluster on the path into difficulty: individuals tend to experience a fall in income, increase their credit limits and take out additional credit. But mortgage-holders are notable for tending to protect their mortgage on the way into distress.

This paper develops a nonparametric methodology for assessing the quality of households’ product search in the (highly heterogeneous) UK mortgage market, requiring no assumptions about preferences over product characteristics beyond non-satiation. Using a uniquely detailed combination of lending transaction reports, product information and credit files for nearly 700,000 UK households that took out a mortgage between January 2015 and July 2016, I am able to conduct pairwise multidimensional comparisons between all products for which each borrower was eligible. This allows me identify choices where the chosen mortgage was strictly dominated on all elements of borrowing cost (interest rates, fees) by another available alternative with the same non-price characteristics. I find that 30% of UK households in the sample chose dominated mortgages, paying £550 ($750) more per year on average as a result. There is strong variation in rates of dominated choices across demographic groups. Going to a lender with whom a borrower has an existing relationship (current account, credit card, etc) is associated with increased probability of making a dominated mortgage choice.

We exploit a change in conduct regulation – the Mortgage Market Review (MMR) in 2014 – to estimate the effects of mandatory advice on the choices of UK mortgage borrowers. To estimate the effect on various aspects of consumer choices, we first construct a unique dataset that combines the entire population of residential mortgage originations from mid-2012 to mid-2016 with credit reference agency data and detailed product data on mortgage contracts. We use machine learning-assisted diff-in-diff matching on the repeated cross-section data to estimate the counterfactual outcome for mortgage borrowers that did not use mortgage advice before the MMR, but were effectively forced to do so afterwards. We find that advice made these borrowers more likely to fix their rates for short periods, choose longer mortgage terms and use an intermediary. We find no effects of mandatory advice on various metrics of borrowing cost. We also use apply a matching estimator to the post-MMR data to describe the effects of using an intermediary on borrower outcomes.