The need for austerity in economies with debt overhangs like Ireland’s has not been disproved by the recent study – Constantin Gurdgiev (interloper)
Since publication of the working paper by Thomas Herndon, Michael Ash and Robert Pollin (HAP), ostensibly subverting the 2010 paper by Carmen Reinhart and Kenneth Rogoff (RR), the Irish Left has, like a small boy who has been told that glucose is good for him, been sugar-high with renewed anti-austerity zeal.
According to the Neo-Keynesianistas, the article by Reinhart and Rogoff ‘Growth in a Time of Debt’ published in the American Economic Review in May 2010, provided the intellectual foundation for the argument that austerity is necessary for countries with public debt in excess of, or near, 90% of GDP. And, they implore, the article has now been demolished by the HAP critique.
In the immediate aftermath of the HAP publication, both new and traditional media were saturated with ‘austerity is dead’ missives from indignant leftists of all shades. The HAP paper’s student co-author became an overnight celebrity.
Alas, the HAP critique of the Reinhart and Rogoff study grossly exaggerated the extent of the errors perpetrated 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 and ignored the large body of academic research on the relationship between public expenditure, economic growth and public debt.
Despite all the hoopla about the HAP study, it confirms the main argument set out in the RR paper, namely that breaching 90% debt/GDP is associated with significantly slower rates of growth. Predictably, the Neo-Keynesianistas ignore this
It is alleged, first, that the authors identified a glaring and undeniable error in the spreadsheet calculation for one of the six main RR findings. This error, unfortunate as it might be, is insignificant to the core conclusions. Correcting for this error changes 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, even to the particular conclusion. A 0.3% swing in growth for an ‘austerity-hit’ economy like, say Ireland or Spain, is negligible. Between 1980 and 2012, the 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 RR paper in computing three of the six core reported results was flawed. In fact, the major discrepancy between the HAP and RR papers is as to which averages matter when it comes to summarising 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.
Between 1945 and 2009 – the period covered by both papers – the US experienced debt-to-GDP ratio in excess of 90% in only four 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 (four years versus19 years) and across countries (the US is structurally robust and much larger; Greece is weaker and smaller). The periods matter: if the US experienced four years of high debt when the global economy was growing slowly, some of the US slowdown would be attributable to global conditions and not to 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: variations in exchange rates, interest rates, and inflation 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 (in excess of 90% of GDP) conditions is exactly the same as it would be in the US. More than that, HAP assumptions also imply that growth volatility around the mean is identical in the US and Greece, despite the fact that smaller economies tend to be much more volatile than larger ones, and that volatility in growth changes over time and across countries. The upshot of the HAP assumption is that the Greek debt overhang is weighted as if it was almost five times more significant than the US’.
In contrast, RR assume that differences across economies and time do matter, and therefore that we should consider separately the average growth rates in the US from those in Greece. The table opposite summarises the differences.
Note that unlike RR, HAP fails to report median results, which are (a) not as different from the HAP mean-based results as RR’s own mean-based results, and (b) were always clearly stated by RR to be the preferred results. The omission of the median findings by HAP is a major one. The difference between the median and average growth rates reported by RR is indeed very sizeable in the case of the countries with debt overhang. This statistically skews 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 RR paper, namely that breaching 90% debt/GDP is associated with significantly slower rates of growth. Predictably the Neo-Keynesianistas ignore this. Uncomfortably for them, the analysis by RR 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 lower levels of economic growth.
The debate re-ignited by the HAP criticism of the RR paper is emblematic of the problem of politicised thinking on the ‘austerity or Keynes’ debate. While 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 borrowing costs for governments, households and companies alike.
The latter points were established for a number of advanced economies across the post-war period in a recent paper from the Bank of Japan (Ichiue and Shimizu, 2013), and by the likes of Laubach; Greenlaw, Hamilton, Hooper and Mishkin; Ardagna; and Baldcacci and Kumar.
The US Congressional Budget Office – hardly a hot house for austerity zealots – clearly shows that for the US the 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 the1973-2012 period to 3.7% of GDP by 2023. By 2018-2020, US Defense and non-Defense discretionary expenditures will be less than the cost of net interest funding.
Fully 65% of all income-tax increases since 2009, including those to be achieved from the forecast increases in economic activity in Ireland up to 2018 will be consumed by the hikes in interest costs on Irish government debt
In the case of another heavily-indebted economy, Ireland, the latest IMF projections show that the interest cost of our debt will rise from 13.3bn in 2009 (2.04% of GDP) to 19.4bn by 2018 (4.6% of GDP). A full 65% of all income-tax increases since 2009, including those to be achieved from the forecast increases in economic activity in Ireland up to 2018 will be consumed by the hikes in interest costs on Irish government debt. The IMF does not publish underlying interest rates and government bond yields. But, 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 between 2013 and 2018. This is optimistic. As I have repeatedly pointed out, we can expect ECB rates to rise to above the 3.1% historical average in the medium-term future. With risk premium broadly consistent with higher Irish debt levels, this may lead to sovereign yields averaging closer to 5% over the 2013-2018 period. In this case, government interest costs could be €12bn or closer to 5.75% of GDP. If this were to occur, the 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 RR 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-à-vis our main trading partners – our high corporate tax rate incentives and low cost basis – drove rapid expansion of our exports. The low-interest-rate 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, while in the 1980s Irish debt levels were flashing red only for government debt, today we have one of the most-indebted private- and public- sector 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 the slower-structural-growth path associated with current economic environment 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.