26/7/17: Credit booms, busts and the real costs of debt bubbles

A new BIS Working Paper (No 645) titled “Accounting for debt service: the painful legacy of credit booms” by Mathias Drehmann, Mikael Juselius and Anton Korinek (June 2017 http://www.bis.org/publ/work645.pdf) provides a very detailed analysis of the impact of new borrowing by households on future debt service costs and, via the latter, on the economy at large, including the probability of future debt crises.

According to the top level findings: “When taking on new debt, borrowers increase their spending power in the present but commit to a pre-specified future path of debt service, consisting of interest payments and amortizations. In the presence of long-term debt, keeping track of debt service explains why credit-related expansions are systematically followed by downturns several years later.” In other words, quite naturally, taking on debt today triggers repayments that peak with some time in the future. The growth, peaking and subsequent decline in debt service costs (repayments) triggers a real economic response (reducing future savings, consumption, investment, etc). In other words, with a lag of a few years, current debt take up leads to real economic consequences.

The authors proceed to describe the “lead-lag relationship between new borrowing and debt service” to establish “empirically that it provides a systematic transmission channel whereby credit expansions lead to future output losses and higher probability of financial crisis.”

How bad are the real effects of debt?

From theoretical point of view, “when new borrowing is auto-correlated [or put simply, when today’s new debt uptake is correlated positively with future debt levels] and debt is long term - features that are present in the real world - we demonstrate two systematic lead-lag relationships”:

• “debt service peaks at a well-specified interval after the peak in new borrowing. The lag increases both in the maturity of debt and the degree of auto-correlation of new borrowing. The reason is that debt service is a function of the stock of debt outstanding, which continues to grow even after the peak in new borrowing.” It is worth noting a well-known fact that in some forms of debt, minimum required repayment levels of debt servicing (contractual provisions in, say, credit cards debt) is associated with automatically increasing debt levels into the future.

• “net cash flows from lenders to borrowers reach their maximum before the peak in new borrowing and turn negative before the end of the credit boom, since the positive cash flow from new borrowing is increasingly offset by the negative cash flows from rising debt service.”

Using a panel of 17 countries from 1980 to 2015, the paper “empirically confirm the dynamic patterns identified in the accounting framework… We show that new borrowing is strongly auto-correlated over an interval of six years. It is also positively correlated with future debt service over the following ten years. In the data, peaks in debt service occur on average four years after peaks in new borrowing.” In other words, credit booms have negative legacy some 16 years past the peak of new debt uptake, so if we go back to the origins of the Global Financial Crisis, European household debts new uptake peaked at around 2008, while for the U.S. that marker was around 2007. The credit bust, therefore, should run sometime into 2022-2023. In Japan’s case, peak household new debt uptake was back in around 1988-1989, with adverse effects of that credit boom now into their 27 years duration.

When it comes to assessing the implications of credit booms for the real economy, the authors establish three key findings:

1) “…new household borrowing has a clear positive impact, and its counterpart, debt service, a significantly negative impact on output growth, both
of which last for several years. Together with the lead-lag relationship between new borrowing and debt service this implies that credit booms have a significantly positive output effect in the short run, which reverses and turns into a significantly negative output effect in the medium run, at a horizon of five to seven years.”

2) “…we demonstrate that most of the negative medium-run output effects of new borrowing in the data are driven by predictable future debt service effects.” The authors note that these results are in line with well-established literature on negative impact of credit / debt overhangs, including “the negative medium-run effect of new borrowing on growth is documented e.g. by Mian and Sufi (2014), Mian et al. (2013, 2017) and Lombardi et al. (2016). Claessens et al. (2012), Jorda et al. (2013), and Krishnamurthy and Muir (2016) document a link between credit booms and deeper recessions.” In other words, contrary to popular view that ‘debt doesn’t matter’, debt does matter and has severe and long term costs.

3) “…we also show that debt service is the main channel through which new borrowing affects the probability of financial crises. Consistent with a recent literature that has documented that debt growth is an early warning indicator for financial crises, we find that new borrowing increases the likelihood of financial crises in the medium run. Debt service, on the other hand, negatively affects the likelihood of crises in the short turn.”

In fact, increases in probability of the future crisis are “nearly fully” accounted for by “the negative effects of the future debt service generated by an increase in new borrowing”.

The findings are “robust to the inclusion of range of control variables as well as changes in sample and specification. Our baseline regressions control for interest rates and wealth effects. The results do not change when we control for additional macro factors, including credit spreads, productivity, net worth, lending standards, banking sector provisions and GDP forecasts, nor when we consider sub-samples of the data, e.g. a sample leaving out the Great Recession, or allow for time fixed effects. And despite at most 35 years of data, the relationships even hold at the country level.”

So we can cut the usual arguments that “this time” or “in this place” things will be different. Credit booms are costly, painful and long term.

2/7/2013: Village June 2013: Real Effects of Government Debt Overhang?

This is an unedited version of my column in the Village Magazine, June 2013 edition.


Ever since the publication of the working paper by Thomas Herndon, Michael Ash and Robert Pollin (HAP) detailing their criticism of the 2010 paper by Carmen Reinhart and Kenneth Rogoff, Irish Left has been abuzz with the anti-austerian sloganeering.

According to the Left’s Neo-Keynesianistas, the article by Carmen Reinhart and Kenneth Rogoff, titled Growth in a Time of Debt and published in the American Economic Review in May 2010 (R&R, 2010) provided the intellectual foundation for the argument that austerity is necessary for countries with public debt in excess of or near the 90% of GDP bound.  And, according to the same Neo-Keynesiastas, the R&R 2010 article has now been demolished by the HAP critique.

In the immediate aftermath of the HAP publication, both the new and the traditional media channels were saturated with ‘the austerity is dead’ missives from angry Leftists of all shades. The HAP paper became the buzzword of the blogosphere, twitter and facebook, and its student co-author became an overnight celebrity.

Alas, the HAP critique of the Reinhart and Rogoff study grossly over-exaggerated the true extent the errors committed by Reinhart and Rogoff. The tidal wave of anti-austerity rhetoric unleashed since the HAP publication has vastly distorted the nature of the original study conclusions and ignored the large body of academic research on the relationship between public expenditure, economic growth and public debt.


Consider the HAP authors’ main charges against the R&R 2010 paper and the case of ‘austerity’ in general.

Firstly, the authors identified a glaring and undeniable error in the spreadsheet calculation relating to one of the six main reported findings contained in the R&R paper. This error, unfortunate as it might be, is neither influential in terms of the original results, nor significant in terms of disputing the core conclusions of the Reinhart and Rogoff body of research. Correcting for this error changes original estimates of the impact of debt on growth by just three tenths of a percent –within the statistical margins of error. In other words, economically, the error was barely significant. A 0.3% swing in growth for an ‘austerity-hit’ economy like, say Ireland or Spain, is indistinguishable from normal volatility in growth rates present in good and bad times alike. Over 1980-2012, standard deviation in real growth in the peripheral euro area states averaged more than nine times the magnitude of the excel error discovered by HAP.

Second, the authors have claimed that the methodology used in the R&R paper in computing three of the six core reported results was flawed. In fact, the major difference between HAP and Reinhart and Rogoff papers is found in the authors differing opinions as to which averages matter when it comes to summarizing countries’ experiences across periods of crises.

The significance of this error can be best understood in terms of a practical example, provided by James Hamilton of the University of California, San Diego.

Since 1945 through 2009 – the period covered by both papers – the US experienced debt to GDP ratio in excess of 90% over only 4 years. In contrast, Greece was in a similar predicament for 19 years. To compare the two countries experiences, one has to deal with the averages across time (4 years vs 19 years) and across countries (the US – with more structurally robust and much larger economy, against Greece – with weaker and smaller economy). Difference between periods matter: if the US experienced 4 years of high debt when the global economy was in slower growth period, some of the US slowdown is attributable to global conditions and had nothing to do with debt overhang. In contrast, if Greece experienced 19 years of debt overhang amidst, say, a robust global expansion, then more of the impact of excessive debt levels can be attributed to internal conditions in Greece. And so on: exchange rates, interest rates, and inflation regimes variations, and other differences between economies at different times – all matter.

HAP assume that the correct way to deal with all these differences is to ignore them completely. Thus, under HAP, the expected growth rate for Greece under debt overhang conditions (debt in excess of 90% of GDP) is exactly the same as it would be in the US. More than that, HAP assumptions also imply that growth rates volatility around the mean is identical in the US and Greece, despite the fact that smaller economies tend to be much more volatile than the larger ones, or that volatility in growth changes over time and across countries. The end result of the HAP assumption is that Greek experience of debt overhang is weighted as if it was almost five times more significant than the US experience.

In contrast, Reinhart and Rogoff assume that differences across economies and time do matter, and this means that we should consider separately the average growth rates in the US from those in Greece.

Table below shows a summary of the HAP results compared to Reinhart and Rogoff results.

Note that unlike Reinhart and Rogoff, HAP fails to report median values, which are (a) not as different from the HAP mean-based results as the R&R mean variables reported, and (b) were always clearly stated as the preferred results by Reinhart and Rogoff. The omission of the median findings reporting by HAP is a major one. The difference between the median and average growth rates reported in the original Reinhart and Rogoff paper is indeed very sizeable in the case of the countries reaching beyond the 90% debt/GDP threshold. This, statistically, indicates that there is a lot of skeweness in the data and suggests that in addition to being associated with lower growth rates, high debt/GDP ratios are also associated with greater risk or volatility in growth.


Despite all the hoopla about the HAP study, it confirms the main argument set out in the Reinhart and Rogoff paper, namely that breaching a 90% bound on Government debt to GDP ratio is associated with significantly slower rates of growth. This is something that the Neo-Keynesianistas are largely ignoring in their calls for scrapping the drive to structurally rebalance fiscal spending and revenue models operating in the countries with already high levels of Government debt. Uncomfortably for Neo-Keynesianistas, the analysis by Reinhart and Rogoff 2010 is broadly and even numerically close to other studies by the two authors which were based on different data and models, as well as to papers from BIS (Cecchetti, Mohanty and Zampolli paper from 2011), ECB (Checherita and Rother, 2010 paper), the IMF (the World Economic Outlook, 2012), and a number of other studies. All of these papers have clearly confirmed that higher debt levels in post-war advanced economies are associated with indisputably lower levels of economic growth.

The debate re-ignited by HAP criticism of Reinhart and Rogoff 2010 paper is emblematic of the problem of politicized thinking on both sides of the austerian-neo-Kenesian divide.  Whilst we do not know much about the causality between debt and growth overall, what we do know is that:
1) Higher debt is associated with lower growth,
2) Higher debt is associated with higher present and future interest rates, and
3) Higher interest rates are associated with higher cost of borrowing for Governments, households and companies alike
The latter points were established for a number of advanced economies and across the post-war epriod in a recent paper from Bank of Japan (Ichiue and Shimizu, 2013), in Vincent Reinhart and Brian Sack 2000 study,  Thomas Laubach 2009 work for the US, Greenlaw, Hamilton, Hooper and Mishkin 2013 paper, Ardagna, 2004, and Baldcacci and Kumar 2010 studies, to name just a few.

The US Congressional Budget Office – hardly a hot house for austerians – clearly shows that US net interest cost of debt financing relative to GDP can be expected to double over the next decade.  This will take net interest cost of funding the US Government debt from 2.2% of GDP in 1973-2012 period to 3.7% of GDP by 2023. By 2018-2020, US Defense and non-Defense discretionary expenditures will be running below those on net interest funding.

In the case of another heavily indebted economy, Ireland, latest IMF projections show that interest on our debt will rise from EUR3.3 billion in 2009 (2.04% of GDP) to EUR9.4 billion by 2018 (4.6% of GDP). Full 65% of all income tax increases since 2009, including those to be achieved from the forecast increases in economic activity in Ireland through 2018 will be consumed by the hikes in interest cost on Irish Government debt. While the IMF does not publish underlying interest rates and Government bond yields assumptions, given the dynamic of debt accumulation, it is relatively safe to assume that the IMF is expecting Irish Government bond yields to average around 4% for 10-year bonds over 2013-2018 horizon. This expectation can be rather optimistic. As I repeatedly pointed out in a number of presentations, we can expect ECB repo rate to rise to above 3.1% historical average in medium term future. With risk premium broadly consistent with higher Irish debt levels, this can lead to sovereign yields averaging closer to 5% over the 2013-2018 period. In this case, Government interest costs can run to EUR12 billion or closer to 5.75% of GDP. If this were to occur, growth in the economy projected by the IMF can fall short of the levels required to deflate our Government debt to GDP ratios.

If neo-Keynesianists think this to be sustainable, we can add the potential impact of higher government yields on cost of funding Irish mortgages and corporate loans.

Another major issue missing in the HAP v Reinhart & Rogoff debate is the question as to whether the aggregate comparatives based on datasets pooling together vastly distinct countries over different periods of time and underlying economic conditions is a meaningful way for looking at the debt overhang problems. In the case of Ireland, consider two sub-periods of high Government indebtedness: the 1980s and the present period. In both, debt/GDP ratios for the Irish Government were running at similar levels. However, the 1990s were associated with Ireland facing an exceptionally robust global demand for its exports. Ireland’s comparative advantage vis-a-vis our main trading partners – our high corporate tax rate incentives and low cost basis – drove rapid expansion of our exports. Low interest rates environment that followed devaluations of the currency has resulted in a series of asset bubbles helping to reduce debt/GDP burden inherited from the 1980s. None of these conditions are present in Ireland today. Lastly, whilst in the 1980s Irish debt levels were flashing red only for Government debt, today we have one of the most-indebted private and public sectors economies in the world.

Which means – in terms of the table above – that we are not starting from a 4%-plus growth benchmark of pre-crisis long term growth trend and we are not heading for a 1.6% median or 2.2% average growth rate in the aftermath of the debt overhang crisis. More likely than not, we are going from a structural growth rate of 2-2.5% pre-crisis to a post-crisis long-term average growth rate of 1%. Whatever Reinhart and Rogoff or HAP aggregates might tell us about the future, it is hardly going to be rosy unless we get our debt and deficits under control and, more crucially, unless we shift our economy from slower structural growth path associated with current economic environment here onto a higher growth path.

How this can be achieved, however, is an entirely different debate from the superficial austerians v neo-Keynesianists ‘to cut or not to cut’ ideological warfare.