Thursday, June 6, 2019

This time might not be different

Estimating the probability of a recession over a short horizon has so far proven to be a challenging task for economists. Each cycle looks slightly different from the previous one and trying to come up with precise indicators of crises leads to either overpredicting them or missing their timing as some risks are underestimated. As the US enters its longest expansion ever, we are back to a discussion on whether there are any reliable indicators that can help us forecast the next turning point. 

Without providing an exhaustive list of all candidates, let me highlight the interaction between three statistical patterns and how they inform us (or not) about the risks ahead: 

Three (related) statistical patterns

1. The Yield Curve tends to invert before a recession.

2. The US does not seem to be able to sustain a low unemployment rate. Once we reach "full employment" (or even before), unemployment bounces back as we hit a turning point. I have written about this pattern in my previous post.

3. No US expansion has lasted more than 120 months. Using the NBER business cycle dates, we are about to enter the longest expansion since their data starts in 1857.

These three statistical patterns are related. As an expansion continues, we see both a gradual decrease in the unemployment rate and a flattening of the yield curve. This should not be a surprise, as unemployment declines central banks raise short-term rates. But what is interesting is that the US (so far) has not been able to reach a state where the yield curve remains flat for a long period of time or, equivalently, the unemployment rate stays low for a number of years. Both the slope of the yield curve and the unemployment follow clear V-shape paths. And this is likely to be linked to the length of the expansion: when the recovery starts both unemployment rates and the slope of the yield curve come down from high levels and as they reach their lowest possible levels, they bounce back setting a limit for how long expansions last. In the current expansion, and after 10 years, even if we started with a high unemployment level (as in 2009), we must be very close to full employment (and the yield curve is flat or inverted).

But aren't these just statistical patterns without an obvious causal argument? Correct, but the fact that this statistical pattern is to robust and consistent means that if the US were to continue its current expansion for a few more years it would have to be that "this time was different". 

Can this time be different?

Could it be that the risks or imbalances that led to previous recessions are either not present or just better managed today? Maybe. It is true that the stock market does not look as expensive as in the year prior to the 2001 recession. It is true that housing markets do not look as expensive as the year prior to the 2008 recession. But we need to remember that in those years we underestimated the relevance of those risks. In 2007 US Federal Reserve Officials praised the resilience of the US financial system to a possible fall in housing prices. Are we failing to see other relevant risks today?

And let's not forget that, even ex-post, some recessions are not clearly preceded by excessive imbalances. For example, the 1990 recession. That recession seems to be more an accumulation of smaller risks combined with geopolitical events (such as the invasion of Kuwait by Iraq). And while some of these geopolitical events are difficult to predict, it is not hard to produce a list of the potential threats the world faces today (from Brexit, to the trade conflicts initiated by the US administration, to the potential instability of the Euro area,...).

In summary, statistical patterns suggest that a recession is imminent. Can this time be different because large imbalances are not present? Maybe. But let's not forget the previous times when we did not see the size and implications of the ongoing imbalances. And let's not ignore the long list of potential risks that could materialize and produce a global slowdown that could easily tilt the US and possibly other countries into a recession. This time might not be different.

Antonio Fatas

on June 06, 2019 |   Edit

Tuesday, March 12, 2019

The 2020 (US) Recession

Summary: This post is based on a research note I wrote asking whether low unemployment is sustainable. The answer is a clear no for the US. Low level of unemployment are good predictors of the tail risk event of a recession, a sharp increase in unemployment rates. These dynamics are related to the build up of financial and macroeconomic imbalances. If this pattern is to repeated, and given the current level of unemployment rate, a US recession must be around the corner. For details on the analysis, the research note including additional results is available on my web site: Fatas (2019).

A few months away from the longest US expansion

The US economy is a few months short of beating the longest expansion ever, which took place from March 1991 to March 2001. As we approach this milestone, there are increasing concerns about the possibility of a recession in the coming years.

Do expansions die of old age? The empirical evidence suggests that this is not the case. There is no clear correlation between the length of an expansion and the probability of a recession (Rudebusch (2016)).

An alternative definition of age

There is an alternative way of thinking about the age of an expansion, in terms of how much economic slack there is left. Expansions are periods where the economy is returning to full employment. As unemployment becomes low and reaches levels around or below the natural rate of unemployment, is it possible to maintain this state for a number of years? Or does “full employment” automatically lead to imbalances that represent the seed of the next crisis?

In the case of the US, history suggests that “full employment” is not a sustainable state and that once we reach such a level a sudden increase in unemployment is very likely. In the figure below I plot unemployment rates around the peak of each of the last five cycles (where zero represents the month the recession started). I plot 5 years before the recession started and 10 months after the recession.

Unemployment Rate Around Recessions (US)
All cycles display a V-shape evolution for unemployment. Unemployment reaches its lowest point around 12 months before the recession and, in most cases, unemployment is already increasing in the months preceding the recession. What is interesting is the absence of a single episode of stable low unemployment (or full employment). It seems as if reaching a low level of unemployment always leads to dynamics that soon generate a recession. Recessions die of old age if “age” is measured in terms of how much economic slack is left. If this pattern was to be repeated, the US must be today very close to an inflection point, a recession.

The result might sound obvious and mechanical: once unemployment rate is low, there is only one way for unemployment to go: up. This is true but what matters is whether a persistent period of low and stable unemployment is possible. In the case of the US the answer is no. One way to visualize this possibility is to look at other countries. A good example is Australia that has recently sustained a low unemployment rate for decades. After a recession in the early 1990s, unemployment increased and then started a decline through a path similar to any US expansions. By the year 2000 unemployment reached a low level that has remained mostly flat for years. In other words, the unemployment rate does not display V-shape dynamics but looks more like an open L-shape.

Unemployment Rate (Australia)

Growth at risk and quantile regressions

We can quantify this intuition by relating this result to an academic literature that analyzes the determinants of the tail risk of unemployment (or GDP) changes. This literature looks at the determinants of worst potential outcomes over a specific time window. Some examples: Cecchetti (2008), Kiley (2018) Adrian, Boyarchenko, and Giannone (Forthcoming). 

Empirically this is done with the use of quantile regressions. In this case we are interested in the tail risk of sharp unemployment increases, which are associated with recessions, and I will capture that by coefficient on the 90th percentile of the distribution in a quantile regression (Fatas (2019)).

The results of such a regression are displayed in the table below. All three coefficients are negative (which is what one would expect as there is reversion to the mean in unemployment rates). But the interesting part is that the size of the coefficient increases as we move from small changes in unemployment to large changes (from q10 to q90). This means that low unemployment rates are particularly good at predicting the tail risk of large increases in unemployment (recessions)

Interestingly, the same phenomenon is not present in other countries (such as Australia). See Fatas (2019) for those results.

Why is full employment unsustainable?

The pattern of US unemployment recessions suggests that low levels of unemployment are a strong predictor of sudden increases in unemployment, associated to crises. We do not observe in the data any sustained periods of low unemployment. But why is low unemployment unsustainable? What leads to a recession? 

The academic literature tends to emphasize two set of variables: those associated to macroeconomic imbalances (such as inflation) and those associated to financial imbalances. Interestingly, the introduction of these variables in the quantile regressions above makes the above effect go away (see Fatas (2019)). In particular, once we control for credit growth, it is not any longer the case that low unemployment is a good predictor of the tail risk associated to recessions (we still observe a reversion to the mean but we do not obtain a larger coefficient for the p90 quantile). 

This result suggests that recessions follow periods of low unemployment because imbalances are built during those years. What is interesting is that the evidence shows that this is always the case, that the US economy has never managed to sustain a low rate of unemployment without generating the imbalances that lead to a recession. If history is an indicator of future crisis, and given the current low level of unemployment, a recession is likely to be around the corner. 


Adrian, Tobias, Nina Boyarchenko, and Domenico Giannone, Forthcoming, Vulnerable Growth, American Economic Review.

Cecchetti, Stephen G., 2008, Measuring the macroeconomic risks posed by asset price booms, Asset prices and monetary policy (University of Chicago Press).

Fatas, Antonio, 2019, Is Full Employment Sustainable?, Manuscript.

Kiley, Michael T., 2018, Unemployment Risk. Finance and Economics Discussion Series, Board of Governors of the Federal Reserve System (US).

Rudebusch, Glenn D., 2016, Will the Economic Recovery Die of Old Age? Federal Reserve Bank of San Francisco Economic Letters.

on March 12, 2019 |   Edit

Monday, December 17, 2018

How low is low for Chinese GDP growth?

The deceleration in the Chinese economy over the last decade has raised concerns about the sustainability of the Chinese economic "miracle". But is that deceleration unusual when compared to other countries? What is to be expected in the coming years?

Economists like to look at emerging markets through the lens of the convergence model (based on the work of Robert Solow). Successful emerging economies are supposed to grow faster than advanced economies and catch up. But as the process of catching up materializes, growth will slow down and over time will approach that of the most advanced economies. How does China compares to other successful emerging economies? It is not easy to find a perfect historical example for China but South Korea comes the closest. It is a successful converging economy in Asia and it is a fairly large economy (unlike Singapore or Hong Kong, two other successful converging economies).

We start by focusing on the period 1980-2018 and use GDP per hour as an indicator of productivity. For each year we compare the initial level of GDP per hour with the growth of GDP per hour over the 5 years that followed. The initial level of GDP per hour is measured relative to the US (as an example of a country close to the technology frontier).

In the context of this period, the deceleration of China makes its growth rate land right at the sample place as the growth rates that Korea had at similar levels of development. The last observation corresponds to the period 2013-2018. Today (2018), China is abut 20% of the US level and if it were to follow the Korean benchmark it would be growing at rates around 6%, very close to current Chinese growth rates. China reached this position after a volatile early decades. Possibly underperforming in the 1980s and over performing in the decade of 2000s when growth passed 10%. 

If we add early decades the comparison becomes much noisier as both South Korea and China had much more volatile, and lower overall, growth rates.

In summary, the deceleration of GDP growth rates in China can be seen as a natural evolution of the economy as it follows its convergence path, in particular if we use recent decades in South Korea as a benchmark. Let's not forget that South Korea is one of the best performer for countries in the range below 50% of the US GDP per capita. So using South Korea as a benchmark we might be providing an optimistic benchmark for Chinese growth.

Antonio Fatás

[Data Source: Total Economy Database, The Conference Board]
on December 17, 2018 |   Edit

Tuesday, November 27, 2018

Global Rebalancing

Prior to the Global Financial Crisis the world economy experienced a period of increasing global imbalances where a group of countries saw their surpluses increase rapidly while, on the other side, a group of countries increased their deficits. These patterns were partly related to the "saving glut" hypothesis put forward by Ben Bernanke to explain the decline in global long-term real interest rates. It was also the case that some of the deficit countries (in particular in the Euro periphery) found themselves in a large crisis after 2008.

This post is an update of the last ten years. Today the world displays smaller imbalances than at the peak of 2008 but what it was more interesting is the extent to which rebalancing had happened between different country groups.

Let's start with the global view. The Figure below shows current account balances as % of world GDP for some regions or groups of countries. Data goes all the way back to 1980 although data is missing for some countries before 1995 (see footnote on data sources).

In the period 1998-2008 (Global Imbalances period) we see the increasing surpluses of oil producer countries, China, advanced Asia (this includes Japan, South Korea, Singapore and Taiwan) and the rest of the world (ROW, many of these countries are emerging markets). The Euro area remained quite balanced (more on this below) and the only deficit country in this figure is the US, absorbing all the surpluses generated by the other countries [of course, the US was not the only deficit country. Some of the Euro area countries had a deficit as well as some of the countries in the rest of the world].

Since 2008 we have witnessed:

  • China is moving fast towards a balanced current account (IMF forecasts suggest that China's current account will be balanced within the next 2 years). 
  • Oil producers have moved towards a balanced CA with small deficits in 2015-2016 as a result of the decline in the price of oil.
  • The Euro area has massively shifted towards a large surplus (the largest among the surplus regions)
  • Advanced Asia has maintained or increased its surplus relative to previous years.
  • The US continues to be the country that absorbs most of the surpluses. The US deficit is smaller than in 2008 but remember this is measured as a ratio of World GDP not US GDP (relative to US GDP the decline would be less pronounced).
  • While in 2008 many emerging markets were savers, in 2018 all the surplus countries are advanced economies (Euro and Asia). Some emerging countries appear under rest of the world as absorbing some of these capital flows.

Some details on the two largest sources of surpluses today. What happened among advanced countries in Asia? The figure below shows that Japan was dominant in this group in the early years. But in recent years the increase in surpluses in Korea, Taiwan and Singapore means the three of them together have a larger surplus than Japan. Overall size is similar to before when measured against world GDP. If we were to measure it against their own GDP we would see an increase in surpluses.

And finally the Euro area. I divide the Euro area into three groups: Germany and Euro Deficit and Euro Surplus. The last two groups are decided by looking at the pattern of the current account of Euro members (excluding Germany) during the 2000-2008 period. Countries that showed consistent deficits are in the first group while the others are in the second one. We can see the large increase in surpluses in Germany in the decade of the 2000s while the group of deficit countries massively increased their deficits. After 2008 we see a fast rebalancing from the deficit countries towards a surplus while Germany maintains (measured as % of world GDP) or increases (relative to its own GDP) the current account surplus.

Antonio Fatás

Data source: Current Account balances (in USD) and World GDP (in USD) from IMF World Economic Outlook database. Data from 1980-1995 is missing for some countries (in particular emerging Asia, including China). During those years they appear as rest of the world.
on November 27, 2018 |   Edit

Wednesday, November 14, 2018

Digital money and payments

New technologies in the financial sector are opening the door for potential disruptions: cryptocurrencies, M-Pesa, WeChat,... Many of them are seen as alternatives to either traditional currencies issued by central banks or to the intermediation role played by commercial banks.

In this discussions, there is often the assumption that "money" and "payments" are features that always come together, they cannot be separated. The confusion originates in the standard definition of money: It is the asset that allows us to purchase goods and services, the "medium of exchange". The ultimate example is physical currency where a piece of paper that says €50 or $100 is both the asset (where the value is being held) and the medium of exchange (the payment vehicle of the payment technology). Transfer of the asset cannot be separated from the "technology" used to make the payment. By giving the note to a seller, you get in return goods and services for exactly that value.

But the moment we think about electronic forms of money, there is a clear separation between the asset and the payment technology. The asset is a balance typically held in a bank (but it can also be in a mobile operator as in the case of M-Pesa). The payment technology is the way I can transfer the value of that asset to someone else. This technology can be a debit card or an NFC chip inside a watch combined with a terminal at a store or it can be a messaging application via an app in your mobile device that connects your balance with the balance of the seller via a network where all institutions operate.

In theory the two features can be treated as separate. A commercial bank can move from a cumbersome and costly payment technology of cheques and inefficient wire transfers that take days to a modern technology where payments and interbank transfers are immediate through a real-time gross transfer system. The nature of money has not changed (the balance in your bank account) but the way money is being used as a medium of exchange (the payment technology) has become much more efficient.

In the real world the two features might come some times together. For example, the case of M-Pesa in Kenya where a mobile phone provider offers a form of money that combines a balance within their systems and a technology to make the payments (via the mobile phone). This is of course more likely to happen in a country where bank accounts are rare so the only way to offer an efficient payment technology was to combine it with a provision of the asset through these balances.

Here is another example where money and payments are being mixed: Christine Lagarde, IMF managing director, speaking at the Singapore Fintech Festival, discussed the benefits of digital currencies issued by central banks (i.e. allowing individuals to hold accounts at the central bank). One of these benefits is "Privacy". Quoting from her speech:

"Consider a simple example. Imagine that people purchasing beer and frozen pizza have higher mortgage defaults than citizens purchasing organic broccoli and spring water. What can you do if you have a craving for beer and pizza but do not want your credit score to drop? Today, you pull out cash. And tomorrow? Would a privately-owned payment system push you to the broccoli aisle? Would central banks jump to the rescue and offer a fully anonymous digital currency? Certainly not. Doing so would be a bonanza for criminals."

In this debate, in order to discuss the benefits and costs of different solutions we also need to separate money from payment technologies. Governments might want to have all the relevant information about the identity of individuals holding money accounts. But they might not care about whether you buy pizza or broccoli; the information about the actual payment. One could imagine a system where the institutions that are holding the assets (money) are highly regulated and compliant with KYC (know your customer) regulations. But the companies that have access to that balance to execute payments do not need to share any information with governments. In fact, we might want them to be required to maintain strong privacy rules regarding the information they collect or sell. No need to create central bank digital currency for all.

Antonio Fatás
on November 14, 2018 |   Edit

Monday, September 10, 2018

Is the Great Moderation back?

The "Great Moderation" was a term used to describe the reduction in business cycle volatility observed in several advanced economies. It started in the mid-1980s and it coincided with the period of time where inflation had successfully brought down to a low level (and remained low and stable since then).

There was a debate about the causes of the Great Moderation. Some put central banks at the center of the phenomenon while others thought good luck was a significant part of the explanation for these benign years. The crisis that started at the end of 2007 represented for some the end of this period and a validation of theories that had seen good luck as the main reason for it. The deep and protracted recession that followed 2007 questioned the idea that business cycles had become less volatile.

But looking at the volatility from today's perspective, 2018, the "Great Moderation" might still be relevant, at least for the US economy. I calculate below a (previous) 5 year standard deviation of real US GDP growth (using quarterly data, growth rates calculated relative to one year earlier).

The data speaks for itself. There is a marked reduction in volatility in the mid-80s that persisted all the way to 2007. Then the increase in volatility is evident, due to the crisis. But in recent years we have seen volatility fall to its lowest levels. This is the result of a a very stable GDP growth and the fact that we are in the second-longest expansion phase the US economy has even seen (10 more months to become the longest one). 

What is interesting is that looking at the whole period 1985 until today, even including the sharp increase in volatility resulting from the global financial crisis, GDP remains much less volatile than in the earlier decades. The Great Moderation seems to be alive in US data.

And here is the same analysis using French data. Similar pattern although because the data start later when volatility was low, it looks more like the exception is the 70s when volatility was much higher than any other period for the French economy. And the surge in volatility after 2007 is stronger partly because the Euro area went through a second recession around 2012.

Antonio Fatás

on September 10, 2018 |   Edit

Wednesday, June 6, 2018

Lost decades: Italy 3 - Japan 0.

The last decade has not been good for many advanced economies. The Global Financial Crisis, a second recession in the Euro area and central banks hitting the zero lower bound have led to disappointing GDP growth rates. But GDP growth rates can be a  misleading indicator about the true performance of different economies. For example, as Matt O'Brien summarizes well, Japan has done much better than what most people believe.

The confusion comes from the fact that there are two forces driving GDP growth. One is the number of hours we work and the other one is how productive those hours are. The absolute number of hours we work is a function of the population of a country. Because of that reason we usually measure GDP per capita growth instead of GDP growth. But this is not enough. Hours per capita can change in response to two forces. First, age matters. Typically engagement in the labor market declines with age, an aging country is likely to have fewer hours per capita. Second, the labor market. Even for groups of the population where labor market engagement should be the highest (prime-age workers) we see interesting variation over time and across countries that reflect on the performance of the labor market.

In this post I judge the performance of countries separating two factors:

  1. Productivity, measured as GDP per hour worked.
  2. The labor market: measured as employment to population for the group of 25-54 years old.

I am just leaving one factor out, which is aging (over which policy makers have little influence except for immigration policies or incentives to increase fertility rates).

Using data from the Conference Board (for productivity) and OECD (for labor market), I have compared the performance of a sample of advanced economies for two periods. 1999-99, which happens to be a pre-Euro period, and 2000-17.

All the numbers are expressed as the value achieved by the variable in the last year as compared to the initial year which is made equal to 100. So an index of 110 means that in the whole period, that variable grew by 10%.

Two caveats: I am measuring performance ignoring initial levels. A country with a very high employment rate is unlikely to increase that rate over time so its performance would seem disappointing. Second, this is about relative performance. Compared to the other advanced economies, how did a particular country perform?

We start with productivity (click for a larger image). Japan does well on both periods. The 4th best performer in the 90s and the 6th one since then. The US climbs up in the post-2000 period ranking. Among the Euro members, Germany loses a few positions with the Euro, same as France, although the change in ranking is smaller. Among the Southern European members, different fortunes. Spain seems to benefit from the Euro membership while Italy and Greece drop to the lowest positions, from already a very low level. For these two countries, performance has been low since 1990 and it got worse since the introduction of the Euro.

There are some countries like the Netherlands that are towards the bottom in both tables. But the Netherlands has one of the highest GDP per hour levels among these countries so it is natural that its growth rate is low. That's not true for Greece or Italy that remain as two of the countries with the lowest GDP per hour level.

What about the labor market? I focus on prime-age workers so that we cannot ignore issues related to aging and effective retirement age. Here Japan is a low performer in the 1990s but then it becomes the best performer in terms of improvements in labor market outcomes for this group. Some of this is related to Abe's policies in particular the increase the female labor force participation. Some Euro members are strong on both periods (Germany and Spain). Italy manages to improve its relative performance in this period although remains in the lower part of the table. Greece, on the other hand, and as a reflection of the large effects of the crisis, becomes the worst performer since the Euro was launched. 

And, interestingly,the US falls to the bottom of the table in the post-2000 period. While productivity-wise the US has done well since the year 2000, the labor market has underperformed relative to the other advanced economies. The UK labor market does well in the post-2000 period and partially compensates for its average productivity performance.

This is still a partial analysis of the data because we are looking at relative performance without considering initial levels (more to come in future posts). But the data so far points to some interesting conclusions:
  • Once the effect of an aging population is removed, Japan has done very well since 1990. It has sustained a good productivity growth while becoming a leader in terms of labor market outcomes in the post 2000 years. No lost decades despite a deflationary environment.
  • Southern Europe has not followed a unique trajectory after the introduction of the Euro. Spain has improved its productivity while maintaining significant improvements in the labor market. Italy was a low performer in the 1990s and since the launch of the Euro its productivity performance declined while its small improvement in the labor market could not do enough to compensate. Italy stands out among the largest advanced economies because performance on all dimensions is low across all decades and this is more remarkable if we consider that both in terms of productivity and in terms of labor markets, its initial level was already relatively low.
  • The US has seen a relative improvement in productivity while the labor market has deteriorated at a rate that only Greece can match.
Antonio Fatas

on June 06, 2018 |   Edit