Revisiting Reinhart & Rogoff after the Crisis: A Time Series Perspective
Summary (4 min read)
1. Introduction
- The authors argue that the field has yet to present a coherent framework with consistent evidence for the existence of a 'debt-threshold' or a strong case for a significant and negative causal link between public debt and economic growth.
- Instead, their findings lend support to those who suggest a reverse causality, where slumps in economic activity are largely responsible for increases in public debt.
2. Relevant Literature
- Within the empirical literature, numerous approaches have been employed in the attempt to shed light on the debt-growth relationship.
- Whilst the authors acknowledge the wider scope of the literature, the purpose of this study is to concentrate on the empirical post-crisis strand which has the 'debt-threshold' hypothesis at its core.
- The authors argue that this particular strand of the literature has yet to present a coherent framework with consistent evidence for the existence of a 'debt tipping point' or a strong case of a significant and negative causal link between public debt and economic growth.
- The authors show this in the preceding paragraphs using the delineating factors of exploratory evidence, correlation and causality, endogeneity, and cross-country heterogeneity.
2.1. Exploratory Evidence
- In their influential study Reinhart and Rogoff (2010) suggest a debt threshold of 90% at which growth is reduced by half for a sample of OECD countries.
- Nonetheless, it would be unjust and rather convenient to hold Reinhart and Rogoff (2010) accountable for the direction into which the public, political or economic debate has been leaning, let alone for the economic consequence of austerity itself.
- Egert (2013) performs a descriptive analysis of the original Reinhart and Rogoff (2010) data, its corrected counterpart (Herndon et al., 2014) as well as a third data set by Egert (2012) where the author had matched government debt data presented in Reinhart and Rogoff (2010) with growth rates from the Barro and Ursula (2010) data set.
2.2. From correlation to causality
- Apart from this algebraic relationship, the endogenous connection between both variables becomes even more illustrative if the authors consider the fact that some policy interventions may explicitly further endogenise both variables:.
- This rational provides a more theoretical argument for a 'reverse causality' link from low growth to high debt where high deficit spending may be one but not the only driving force of crumbling GDP growth rates.
- As shown in Baek and Brock (1992) , such parametric and linear Granger causality tests are found to have low power against certain non-linear alternatives.
2.3. Endogeneity in a cross-country setting
- Panizza and Presbitero (2014b) instrument the debt-to-GDP ratio with the valuation effect of debt held in foreign currency and exchange rate movements.
- Their exogenous instrument for country i is the averaged debt-to-GDP ratio of all remaining (12−i) countries; a proposition which may cast doubts in terms of validity: Checherita-Westphal and Rother (2012) also employ endogenous instruments to combat the challenges of endogeneity by means of a GMM estimation framework.
- Unfortunately, the inclusion of lagged, endogenous instruments may not adequately address the endogeneity problem in this particular case as this method was initially designed with microeconomic purposes in mind.
- To summarise, empirical efforts to unfold the debt-to-growth relationship through a cross-country framework are challenged by their capability to adequately control for model simultaneity.
2.4. Heterogeneity
- Important implications arise around the issue of cross-country heterogeneity.
- Particularly for a set of homogeneous and strongly interconnected entities, cross-sectional dependence may severely bias estimation results.
- The importance of accounting for cross-country heterogeneity is also stressed by Chudik et al. (2015) who provide evidence of cross-sectional heterogeneity of the 'debtthreshold' hypothesis in a dynamic, heterogeneous panel set-up.
- The authors provide threshold estimates of between 60 to 80 percent for the full sample as well as 80 percent for the advanced economies and between 30 to 60 percent for developing countries respectively when not accounting for unobserved common factors.
- Their cross-sectionally augmented models fail to reject the hypothesis of no simple debt threshold.
2.5. Summary
- In this section the authors have surveyed a wide and contradictory field of academic research concerned with the importance of public debt for economic activity.
- The authors have focused on the most consistent and well developed body of 'debt literature' in advanced economies in terms of modelling choices.
- The authors reading of the literature is that the debt-growth relationship is both more complex and heterogeneous, when considered across countries and time periods, than initially assumed.
- What is more, it also raises the question as to whether gross debt is indeed the correct measure for this investigation in the first place, because of its inherent inability to consider the cross-sectional heterogeneity of national debt compositions.
- One interesting characteristic of the debate, to date, is that as the argument develops over time the consensus appears to be reversing, with later studies tending to disagree with the existence of the debt threshold.
3. Exploratory Data Analysis: A Time-series Perspective
- In this section the authors revisit the central claims made in Reinhart and Rogoff (2010) by employing a novel exploratory approach to visually disentangle the linkage between economic growth and public debt.
- The authors do this by using the corrected Reinhart and Rogoff (2010) data set following Herndon et al. (2014) which they contrast with up-to-date and higher-frequency OECD data.
- The authors offer a deliberately uncomplicated contradiction to the findings of Reinhart and Rogoff (2010) and the earlier supporting literature, by employing basic visual and exploratory techniques, in a similar way to Reinhart et al (2012) , allowing us to highlight channels and directions of causality, that might remain difficult to disentangle in more sophisticated empirical techniques.
- In doing so; the authors use an approach that they hope will facilitate wider access to this important debate, beyond the field of economics.
- The authors goal is to develop insights regarding the time-dependent causality pattern of debt and growth, their structural nature as well as any potential threshold effects of debt by comparing the patterns of the same set of countries and periods at different frequencies.
3.1. Inter-temporal Linkage of Debt and Growth
- In order to evaluate the relationship between public debt and time the authors introduce three regimes for debt and growth which they define based on the findings in the literature:.
- There also seems to be little visual evidence to suggest that the absorption of the financial crisis reflected by the decline in the growth rates seem to be associated with the debt regime of the corresponding countries in any way:.
- The authors try to address this issue in Figure 3 where they again use Reinhart and Rogoff (2010)'s annual data set and plot the levels of the debt-to-GDP ratio against time.
- If high public debt is causing low economic growth, the authors would assume to see more medium and low growth regimes associated with countries with higher debt levels.
- Rather than being typically present for extremely high levels of the sovereign debt ratio, periods of low economic growth seem to occur at similar time periods across countries (and have become more frequent in the 1980s to mid-1990s).
3.2. Non-linearities and Thresholds
- The authors acknowledge the argument that even if previous debt levels might not have been detrimental in causing low economic growth, as suggested in their analysis, this does not mean that sovereign debt induced economic slowdowns are an impossibility.
- This result is intuitive and is in line with their previous observations:.
- Crises can typically be characterised by a sharp drop in average growth rates and seem to generally lead to a debt build-up afterwards.
- Nevertheless, growth figures are nowhere near the negative aggregates reported in Reinhart and Rogoff (2010) for the 90% threshold, of which there appears to be very little evidence.
4. Conclusion
- The continuing debate over the imposition of measures to deal with sovereign indebtedness, since the financial crisis, has become centred on the seminal contribution by Reinhart and Rogoff (2010) .
- Since the publication of various inaccuracies in methodology pointed out by Herndon et al. (2014) amongst others, the argument has continued unabated.
- Secondly, after the financial crisis, it is evident that the number of higher debt regimes has exploded; offering support for reverse causality of debt caused by economic slumps.
- Thirdly, it seems that periods of low growth tend to be clustered around common time periods for many countries and not bound to country samples displaying particularly high levels of debt, thereby offering further support for reverse causality.
- The authors demonstrate that the average negative relationship between economic growth and the debt-to-GDP ratio strongly depends on the time dimension of consideration, and that accounting for business-cycle effects and short-term volatility accounts for much of the negative variation in growth figures.
Did you find this useful? Give us your feedback
Citations
37 citations
11 citations
10 citations
4 citations
Cites background from "Revisiting Reinhart & Rogoff after ..."
...Indeed, Amann and Middleditch (2015) show recently that there is no support for the view that higher levels of debt cause reductions in economic activity (they did not include currency and stock market crises either)....
[...]
3 citations
References
292 citations
268 citations
212 citations
197 citations
187 citations
Related Papers (5)
Frequently Asked Questions (14)
Q2. What future works have the authors mentioned in the paper "Revisiting reinhart & rogoff after the crisis: a time series perspective" ?
When analysing the inter-temporal linkage between growth and debt, their time series analysis provides crucial evidence on the much debated possibility that growth can be endogenous on the level of external public debt. The authors build this argument, firstly, on the observations that the times series plots for growth in GDP, over the last six decades, reveal a downward trend over the sample ; a fact which could be the actual explanation for any negative correlation between debt and growth. Thirdly, it seems that periods of low growth tend to be clustered around common time periods for many countries and not bound to country samples displaying particularly high levels of debt, thereby offering further support for reverse causality. In addressing the question of the existence of a debt threshold, the authors find no evidence that countries will experience significant reductions in GDP growth after surpassing a certain percentage of the debt-to-GDP ratio.
Q3. What is the immediate issue arising from the article?
An immediate issue arising is the obvious endogeneity problem resulting in an estimation bias of(2)which the authors derive in Appendix C.
Q4. What is the significance of cross-sectional dependence?
Particularly for a set of homogeneous and strongly interconnected entities, cross-sectional dependence may severely bias estimation results.
Q5. How many countries have used the Granger causality test?
In a sample of 16 OECD countries spanning 30 years Puente-Ajovín and Sanso-Navarro (2015) apply a panel bootstrap Granger causality test developed by Kónya (2006) which accounts for cross-country heterogeneity as well as cross-sectional dependence.
Q6. How does the authors find evidence of a common cross-country debt threshold?
Using a dynamic Common Correlated Effects (CCE) Mean Group estimator which allows for varying country-level coefficients they find no empirical evidence of a common cross-country debt threshold.
Q7. What is the purpose of this study?
Whilst the authors acknowledge the wider scope of the literature, the purpose of this study is to concentrate on the empirical post-crisis strand which has the ‘debt-threshold’ hypothesis at its core.
Q8. What makes the threshold argument problematic in the eyes of the authors of this study?
What makes the threshold argument problematic in the eyes of the authors of this study is that through their ad-hoc selection rule, Reinhart et al. (2012) identify 26 debt-overhang periods for only 13 out of their sample of 22 advanced countries and furthermore note that ‘[...] many debt overhangs result from costly wars.
Q9. What is the way to address this issue?
A straight-forward way common within the literature to address this issue is by treating both GR and D as endogenous by means of a vector autoregressive (VAR) model of the form(3a)(3b)where (pGR, pGR) denotes the selected lag length for the endogenous variables.
Q10. What is the main reason why the authors fail to reject a non-Granger causality?
Another study which analyses a Granger causal relationship is Irons and Bivens (2010) who study the debtgrowth nexus for the case of the USA and fail to reject non-Granger causality going from debt to growth but finds significant evidence of a reverse causality link from growth to debt.
Q11. What does the author say about the relationship between debt and growth?
they report that the association of debt and growth tends to weaken for higher levels of debt, particularly when controlling for peer country growth rates.
Q12. How does Baum et al. (2013) find evidence of a significant increase in the debt?
As discussed before, Baum et al. (2013) find evidence of a considerable increase in the significant debt threshold (from 66 to 72 and 96 percent for the non-dynamic and dynamic panel respectively) if the crisis years 2008 to 2010 are added to the analysis.
Q13. How do they approach the challenges of endogeneity?
Checherita-Westphal and Rother (2012) also employ endogenous instruments tocombat the challenges of endogeneity by means of a GMM estimation framework.
Q14. What is the main result of Reinhart and Rogoff’s study?
More precisely, their ‘[...] main result is that whereas the link between growth and debt seems relatively weak at ‘normal’ debt levels, median growth rates for countries with public debt over roughly 90 percent of GDP are about one percent lower than otherwise’ (Reinhart and Rogoff, 2010, p. 573).