Recession and recovery signposts
VoxEU Column Business Cycles COVID-19 Monetary Policy

Supply shocks, monetary policy, and scarring

There are heightened concerns that advanced economies are on the brink of a recession, but even the pessimists generally agree that economies will eventually converge back to trend. This column casts doubt on this view. Using a panel of 24 countries from 1970 onwards, the authors find that there is a significant tipping point in recovery dynamics, which depends on the size of the initial contraction: severe contractions have highly persistent effects, whereas smaller contractions do not. This finding holds regardless of whether the contraction follows aggressive monetary tightening to combat high inflation, energy price disruptions or other factors (including financial crises).

The current outlook for economies globally is bleak. Headline inflation rates around the world have soared following supply-side disruptions from the war in Ukraine, previous Covid-19 lockdowns, and possible further lockdowns in China. As central banks have pivoted to a more aggressive tightening phase, the question of whether they can manage soft landings or whether there will be recessions has been hotly debated (e.g. Boissay et al. 2022, Summers 2022, Yellen 2022). 1 But despite major uncertainties ahead, private and public forecasters generally agree that the economy will revert back to trend in the medium run. One reason for this is that full recoveries are essentially baked into standard macro models qua assumption.

The data do not support such confidence in the stability of the medium-term outlook. Our research with Xiaochuan Xing (Aikman et al. 2022) highlights the fact that deep contractions more often than not leave long-term economic scars. It is well known that such scarring can occur following recessions caused by financial crises, such as that of 2007-09 (e.g. Cerra and Saxena 2008). What is less well known is that scarring also occurs after deep recessions associated with other factors, such as monetary policy tightening or oil price disruptions. 2 As we discuss below, this has important implications for policy.

The scarring effects of deep recessions

Our results are based on a simple, intuitive test, which is illustrated in Figure 1.

We first bucket together contractions of similar magnitudes in our dataset, starting with the weakest 5% of outcomes across countries. For these most severe contractions, we compare 10-year ahead growth rates calculated immediately prior to the contraction with those from all other points in our sample. If real output eventually recovers its previous trend, there should be no statistical difference between these two long-term growth rates on average. However, if contractions cause scarring, long-term growth rates taken from their eve will be lower than the rest.

We then repeat the process for successively milder outcomes in 5% intervals until we reach the median. 

Figure 1 Detecting scarring effects with long-run growth rates: An illustration

Figure 1 Detecting scarring effects with long-run growth rates: An illustration

Note: Illustration of the impact of alternative shocks at time t0 on multi-period growth rates. From the perspective of t0, a permanent shock (left panel) of size -d reduces the h-quarter ahead growth rate by -d/h (blue dotted lines). In contrast, if the shock is transitory, the average growth rate returns to the trend for large enough horizons (right panel).

Our sample contains 19 advanced economies and five emerging market economies (EMEs) from 1970 to the present. 3 We remove long-run growth trends and normalise growth rates by country-specific means and standard deviations to make sure that they are comparable across countries.

Figure 2 shows the scarring effects of deep contractions. The dark red lines show the difference between the average 10-year ahead real GDP growth at the onset of the contraction and the rest of the sample. We calculate these differences for contractions of different sizes as described above (shown on the x-axis). Grey areas are confidence intervals.

The worst 20% of contractions in terms of severity have highly persistent costs whose effects are visible in the level of GDP a decade hence. But the magnitude of the scarring increases more than proportionally with the size of the initial contraction. For the deepest contractions, i.e. those in the worst 5%, GDP growth in the decade that follows is substantially below the full sample mean. This translates into a persistent loss of 4.25% in the level of real GDP for an average economy. 4

We also show in the paper that while the specific tipping point may vary with the test specification, the main takeaway that only severe contractions have scarring effects is very robust across samples, time periods, detrending methods, and approach used to identify contractions.

Figure 2 Only severe contractions have scarring effects  

Figure 2 Only severe contractions have scarring effects  

Note: Difference in mean 10-year growth rates following contractions of different severity versus that calculated using all other points in the sample. The solid lines show points estimates of this difference for contractions in the severity buckets indicated on the x-axis. The shaded areas are 95% confidence intervals. Severity buckets are based on distribution of annual real GDP growth outcomes starting with the weakest 5% outcomes (“most severe”) to then look at successively milder outcomes in 5% intervals until the 45th to 50th percentile (“mild”). The y-axis is standard deviations of 10-year real GDP growth.

A question of particular interest at this juncture is whether all large contractions come with long-term losses, or whether only certain types of contractions do. To analyse this, we focused on the 5% most severe contractions and classified them into one of the following four categories: (1) banking crisis; (2) restrictive monetary policy to combat high inflation; (3) oil shocks; and (4) other. In total, out of 198 large recessions we classified 100 as banking crisis-driven events, 51 as monetary policy-driven, 19 as driven by oil shocks, and 9 as having other causes.

Using this classification, we find that the specific cause of the contraction does not matter much for our results. 5 When conditioning on each type of contraction separately, we find long-run effects of roughly similar magnitude for banking crisis and those related to monetary policy tightening and of greater magnitude for oil shocks (Figure 3). Only “other” contractions, for which it is difficult to identify a single clear cause, are not associated with any significant long-lasting effects.

Why should such permanent losses occur? Likely mechanisms historically have included slower technology growth due to lower investment in R&D, bankruptcies of highly productive start-ups, or the erosion of human capital from long spells of unemployment. At this juncture, an additional mechanism could be the process of deglobalisation and fragmentation kickstarted by Covid and the Russian invasion of Ukraine. 6

Figure 3 Long-term growth effects of severe economic contractions by proximate cause

Figure 3 Long-term growth effects of severe economic contractions by proximate cause

Note: This chart shows mean differences in 10-year growth rates after annual growth contractions in the worst 5% in terms of severity compared to the rest of the sample. The x-axis shows the different recession types. The y-axis shows standard deviations of 10-year growth rates.

What does this all mean for policy?

For one, our results put renewed emphasis on Olivier Blanchard’s assessment after the Great Financial Crisis that “authorities should make it one of the major objectives of policy—macroeconomic, financial regulatory, or macroprudential—to stay further away from the dark corners” (Blanchard 2014: 30), where he defines dark corners as those states of the world when the economy malfunctions badly and non-linearities kick in.

But as several bad shocks have already occurred, a key question is whether policymakers have sufficient tools to avoid such a dark corner materialising, especially if they want to combat rising inflation at the same time. Many current forecasts anticipate that central banks will eventually deliver full recoveries. As such, long-term effects of the current predicament may be limited, with economies around the world eventually reverting to the previous trend.

The situation is frail, however. It is, for instance, not inconceivable that supply disruptions could worsen as the war continues, or that Covid could surge again and hamper the global economy. Weighing these trade-offs is beyond the scope of our paper. But what is very clear from our analysis is that relying on standard macroeconomic models that assume the economy will inevitably recover the previous trend after a shock can be misleading, especially at times when the stakes are highest. And while there are many risks ahead, relying on wrong models is certainly one risk that policy makers can avoid.

Authors’ note: The views represent those of the authors and not necessarily those of the Bank for International Settlements or the Bank of Finland.


Aikman, D, M Drehmann, M Juselius and X Xing (2021), “The scarring effects of deep contractions”, BIS Working Paper No. 1043 and Bank of Finland Research Discussion Paper 12/2022.

Bank for International Settlements (2022), Annual Economic Report 2022.

Bartholomew, L, and P Diggle (2021), “The lasting impact of the Covid crisis on economic potential”,, 21 September.

Blanchard, O (2014), “Where danger lurks”, Finance and Development, p 28-30.

Blanchard, O, E Cerutti and L Summers (2015), “Inflation and Activity – Two Explorations and their Monetary Policy Implications”, NBER Working Paper No. 21726,

Boissay, F,  F De Fiore and E Kharroubi (2022), “ Hard or soft landings”, BIS Bulletin No. 59.

Borio, C, and E Kharroubi (2016), “Financial cycles, labour misallocation, and economic stagnation”,, 14 April.

Cerra, V, and S Saxena (2008), “Growth dynamics: The myth of economic recovery”, American Economic Review 98:(1) 439–457.

Cerra, V, A Fatas, and S Saxena (2020), “The persistence of a COVID-induced global recession”,, 14 May.

Claessens, S, M A Kose, and M E Terrones (2012), “How do business and financial cycles interact?”, Journal of International Economics 87(1): 178–190.

Jorda, O, M Schularick, and A Taylor (2013), “When credit bites back,” Journal of Money, Credit and Banking 45(2): 3–28.

Malmendier, U, and L Shen (2021), “Scarred consumption”,, 15 March.

Portes, J (2020), “The lasting scars of the Covid-19 crisis: Channels and impacts”,, 1 June.

Reinhart, C M and K S Rogoff (2014), “Recovery from financial crises: Evidence from 100 episodes,” American Economic Review 104(5): 50–55.

Summers, L (2014), “U.S. Economic Prospects: Secular Stagnation, Hysteresis, and the Zero Lower Bound”, Business Economics 49(2).

Summers, L (2022), "We have to be prepared for a recession", June 13, 2022.

Yellen, J (2022), Transcript of Secretary of the Treasury Janet L. Yellen’s Press Conference, 28 July.


  1. For the US, the latest Consensus Forecast for real GDP growth is on average 0.5% for next year ranging from -0.8% to 2%.
  2. See also Cerra et al. (2020), who discuss the potential for scarring effects to occur in the aftermath of the Covid recession.  Scarring is related to, but conceptually distinct from, the concept of secular stagnation, whereby long-term trends in the economy such as lower population growth can lead to a persistent decline in the equilibrium real rate of interest (Summers 2014).
  3. The following countries are included: Australia, Austria, Belgium, Brazil (EME), Canada, Denmark, Finland, France, Germany, Italy, Japan, Korea (EME), Mexico (EME), the Netherlands, Norway, New Zealand, Portugal, Sweden, Singapore (EME), South Africa (EME), Spain, Switzerland, the United Kingdom, and the United States. Real seasonally adjusted GDP series are available from 1970Q1 for all countries, except Singapore (1975Q1) and Brazil (1980Q1). Data are taken from national sources.
  4. Our results are in line with the findings of, for example, Blanchard et al. (2015), who use a different approach to investigate hysteresis effects.
  5. This result broadens that of other papers which find, specifically, that banking crises have permanent effects on output (e.g. Cerra and Saxena 2008, Claessens et al. 2012, Reinhart and Rogoff 2014, Jorda et al. 2013).
  6. There is an extensive literature on the channels via which scarring effects can occur.  For example, Portes (2020) and Bartholomew and Diggle (2021) discuss relevant channels in the context of the Covid-19 recession.  Borio and Kharroubi (2016) present evidence that resource misallocation in a credit boom can lead to persist aggregate demand weakness in the bust.  And Malmendier and Shen (2021) articulate a novel channel via which deep recessions lead to greater pessimism about future prospects and hence lower future consumption.