Household Debt in the UK

Yesterday I argued that if you want to discuss what changes to household debt might mean, you need a decent framework for assessing the fluidity of the microstructure of debt. To illustrate I set out two hypothetical countries with identically modest aggregate household debt loads – but with different sensitivities to changes in interest rates, and differing levels of financial stability. I called them Debtzania (where debt is concentrated and debt-to-income ratios for debtor was high) and Feudaland (where debt is thinly spread and a few households are creditors).

Which of these economies does the UK most resemble?

Using the NMG/ Bank of England survey data we can examine where it is in the income distribution that UK household debt falls.

Table 2 shows that most debt is owed by those in the top three deciles of the income distribution, and on sums below £175k. The mortgage-dominated nature of this debt and the untroubling LTVs with which it is associated may explain it to some as both rational and desired.

Table 2: Where household debt sits, by pre-tax income decile and total debt cohort

We can perform the same analysis for deposit balances held among our sample (knowing that at the macro level, total monetary assets and total monetary liabilities in the household sector are roughly the same), the output of which is shown in Table 3. It is perhaps not surprising that most of the money is owned by members of the upper half of the income distribution.

Table 3: Where household deposits sit, by pre-tax income decile and total deposit cohort

We could treat each income decile as a separate sector and look at the transfer of net monetary claims between them (in the form of savings and borrowings). This is shown in Table 4, and the data suggests that the flow of savings is on a net basis from lower income cohorts to upper income cohorts, typically to deliver mid-sized mortgages (we could, of course, reverse the causality in this description). This concentration of debtors might sound a bit more Debtzanian than Feudalish.

Table 4: UK households by pre-tax income decile and total debt cohort, estimated net saving/ borrowing flow

But debt sustainability is only really an issue for those in debt, and will correspond to their ability to service debt. And so in Table 5 we examine how debt-to-income ratios vary across the our matrix of UK household debt and income distribution. After all, small cohorts of high debt-to-income ratios was the thing identified as problematic when looking at otherwise untroubling Debtzanian aggregate debt statistics. Here we can see that debt-to-income ratios look pretty high for almost all debtors, but least troubling for highest income households where most nominal debt resides.

Table 5: UK household debt-to-income ratios, by pre-tax income decile and total debt cohort

And so the problems for policymakers seeking some rules of thumb to estimate sustainable or unsustainable levels of household debt appear twofold.

First, there is a data issue. Our sample size of around 6,000 households is small (as evidenced by @jmackin2’s outrage yesterday), even before we start to exclude households that refuse to give data in which we have interest. For the purpose of this blog I have used a snapshot, but I think that we need to compare datasets on a longitudinal basis. When the numbers on the matrix change, we need to understand whether this the result of movement on the income dimension or debt dimension.

Second, we have a question over the appropriate unit of analysis. In discussing Feudaland and Debtzania we argue against this being the overall economy, but as suggested by a comparison between the matrices of debt and income distribution on the one hand, and deposits and income distribution on the other, splitting the population into income deciles takes us not much closer to the issue at hand. There are meaningful flows between debtors and creditors within members of any given income decile. The appropriate unit of analysis is perhaps the household. But that statement seems a bit nuts.

The chart below looks at individual households in the NMG survey, rather than treating them as aggregated cohorts. It’s messy, but maybe messy is what works.

Putting aside human concerns for families struggling with the very real problems attached to sever indebtedness, why should policymakers care?

FPC members care because widespread unsustainable household debt can lead to problems for financial stability. But, through this lens, problems in debt service among low income households with small but unsustainable debt balances are just not worthy of deep interest: they are insufficiently likely to cause systemic financial problems.

MPC members care because changes in appetite for debt might tell them something about the location of the neutral real interest rate. Through this lens it might seem that it doesn’t matter what individual households are doing, but does matter what they collectively do. And I guess that if this is true, changes in the optimal level of debt which can only be inferred from an examination of its microstructure will be an important in any decision as to the wisdom of ‘leaning against the wind’. And they will care because cashflow impacts for indebted households following changes in interest rates are part of the transmission mechanism.

What to do?

Despite data shortcomings, there are some suggestions that we can draw out. First, in considering changes to household debt, policymakers should do well to treat aggregate debt statistics with caution. Having an idea as to whether they are dealing with Feudaland or Debtzania, and the degree to which there is stability in the microstructure of the debt distribution seems important. Second, examining the entire distribution of nominal debt-weighted debt-to-income ratios (and changes thereof) appears to be a step towards more capturing more meaningful household debt characteristics, changes in household behaviour and systemic debt problems.

Reverse ferret on household debt

A couple of blogs ago I posed the question as to how the sustainable level of household debt might be identified for an economy. I then reckoned that there might be a decent 30k ft way of addressing this question. This post serves as a bit of a reverse ferret. I’m not saying that the previous approach was wrong, but after some back-to-basics thinking I understand that its rightness is overly-dependent upon the stability of something that it may or may not be wise to assume. In my defence, I reckon *every* macro approach relies on this stability (comments section please to correct me). But I take little comfort in expressing a popular view based solely on its popularity.

Debt is an entirely distributional issue. Evenly distributed in a closed economy, no level of household debt is unserviceable or unsustainable. But evenly distributed, debt is also functionless. Every sustainability question rests upon the microstructure of debt loads across different households. And so, from a macro perspective, debt is tricksy to say the least.

This all becomes clear in an example. Let’s imagine two super-simple closed economies with no sectoral flows: the Kingdom of Feudaland and the Republic of Debtzania.

The Kingdom of Feudaland consists of 10m households, each of them with perfect equality of employment income (at £20k per annum), making its GDP worth £200bn per annum. Household debt sums to £20bn, and interest rates are 15%. Debt is owed by 9.625m households to the other 375,000 remaining households (evenly). Interest costs are eminently serviceable (£312 per annum per household in transfers from the debtors to the creditor households, each of whom receives a handsome £8,000 per annum in interest).

The Republic of Debtzania also consists of 10m households, each of them with perfect equality of employment income (at £20k per annum), making the GDP worth £200bn per annum. Household debt also sums to £20bn, and interest rates are also 15%. Debt is owed by 375,000 households to the other 9.625m households (evenly). Servicing this debt costs the debtor households £8,000 per annum. The creditor households each receive a modest £312 in interest income on their savings.

Table 1: Feudaland and Debtzanian debtmetrics

The aggregate debtmetrics for Feudaland and Debtzania are exactly the same (Table 1). At 10%, the aggregate household debt to income ratio looks pretty low, as do aggregate interest payments as a proportion of GDP at 1.5%. But the microstructure of the debt distribution is what matters in assessing the sustainability or otherwise of these two economy’s debt loads.

Feudaland, despite its wealth inequality, looks to be a beacon of financial stability. The debt-to-income ratios for debtors are pretty low at 10.4% of income, and servicing ongoing interest costs is universally manageable, absorbing 1.6% of income. Feudaland’s wealth inequality might eventually prove politically unsustainable, but it would be hard to call it financially unsustainable, even in an environment where interest rates doubled.

Despite exactly the same aggregate debt-metrics, Debtzania looks much more financially fragile than Feudaland. The majority of its citizens are savers, but they have amassed claims on the small minority of financially-fragile debtors who must pay out 40% of their incomes in interest payments to keep out of default. Debtzania debtors’ debt-to-income ratios are individually high at 267%, and it is probably fair to say that a doubling of interest rates might bring meaningful problems to Debtzania’s financial system (with associated morality stories).

Now imagine being a monetary policy maker in each of the two countries. At what point should you worry about changes to aggregate household debt, and at what point could you infer that household debt in your country is too high? If a policymaker looked to their countries’ respective histories for clues they would, in so doing, be making the assumption that the microstructure of debtor-creditor relationships would be relatively static (or at least semi-stable) over time.

With a static microstructure of debt in Feudaland we might expect to see household default rates relatively unresponsive to wild changes in aggregate debt loads or interest rates. Policy makers might draw the lesson that household debt – at only 10% of GDP – is far too low to worry about, and that rising debt could be a sign that households were moving towards their optimal debt load (and so should not be met by any ‘leaning against the wind’). One could even imagine the cultural trope of creditworthiness being assigned to Feudalanders (backed by the empirical evidence of minuscule historical credit losses).

With a static microstructure of debt in Debtzania, we might by contrast expect to see household defaults rise and fall with both real (and nominal) interest rates and debt-loads. Those 375k households who do the borrowing may – if cross-country evidence holds in our mythical land – do a disproportionate amount of the spending in the economy. And so monetary policy might prove a particularly speedy macro tool for demand management: introducing higher and lower pain thresholds to debtor households and transferring higher or lower amounts of debtor household incomes to creditor households with lower marginal propensities to spend. It would be unsurprising if Debtzanians acquired a reputation of being quick to default; concerned politicians might encourage them all to save more.

In short, if the structure of the distribution of debt was utterly unchanging in both countries, policymakers might reasonably rely on each country’s individual history of debtmetrics to determine policy pinch-points.

But the usefulness of any historical comparisons (in terms of aggregate debt-to-income ratios or debt-service ratios) rests on this microstructure of debt being pretty concrete. The more changeable the microstructure, the less useful any historical comparisons.

Imagine again that the microstructure of debt distribution in Feudaland shifted to match the distribution in Debtzania, perhaps as its demography shifted through population ageing and associated changes in savings habits. A policymaker following only aggregate debtmetrics might infer that the change in default patterns was associated with some cultural shift in the population. When in fact, aggregate debtmetrics had instead done a good job at masking this demographic sea change.

It is not clear that I, or anyone else who makes macro comments on matters of household debt, has a decent framework for assessing the fluidity of this microstructure.

In the next post I look at the data that I can see readily available for the UK.