From Main St. to Wall St.: The long run

What is the long-run relation between economic activity and the stock market? From Main Street to Wall Street analyzes, inter alia, this question. Here, I present some of the conclusions.

As mentioned in my previous post (link), From Main Street to Wall Street describes how economic activity influences financial markets, in particular stock markets. It distinguishes between the long-run and the shorter-run relation. The “long run” refers to decades, even centuries. The short run refers to months, potentially a few years.

This post presents some of the book’s conclusions with respect to the historical long-run relation between the economy and the stock market. This is both an intriguing and fascinating topic, and the conclusions are not always as one would have guessed a priori.

The relation between stock returns and economic activity

The intuition of many people is that stock returns are high when economy growth is high. This is a good starting point. I emphasize in my book, though, that things are not always as simple as that. The long-run relation between economic activity and stock markets illustrates this.

A key observation is that, in the long run, stock returns and economic growth do not go hand in hand, generally. This is counterintuitive for many people, and thus important to emphasize. People have this tendency to believe that when the economy is doing well, firms prosper, which should lead to high stock returns, right? I emphasize and demonstrate that – in the long run – the story is more subtle.

There are several ways to illustrate this.

In those parts of the book that deal with the long run, I typically examine data spanning the past 100 to 150 years. Let’s start with a simple illustration. The average growth rate of what is still the world’s largest economy, the US economy, has been two percent per annum over the past 150 years (average annual growth rate of US real per capita GNP). Has the average annual real return from the US stock market over the same period also been 2%? No(!), for sure not. The average annual real US stock return has been close to 7% over the same period.

A five percentage-point difference (7% – 2%) might not seem like a lot, but compounded over 150 years it is enormous. To illustrate, imagine you (or one of your great-grand parents) invested one US dollar in the US stock market in 1871. Imagine also that you, since then, have reinvested all dividends. I.e., you invested one USD in the stock market many years ago, held on to the investment, and reinvested all dividends. The cumulative value of this investment strategy would have turned into app. USD 15,000 after 150 years. This is even in real terms, i.e. after taking account of inflation. The long-run performance of the US stock market has been truly amazing. USD 1 has turned into USD 15,000.

What about the performance of the economy? If we normalize economic activity to USD 1 in 1871, i.e. similar to our investment of USD 1 in 1871, then the cumulative value of US economic activity (real capita GNP) is USD 17 after app. 150 years. Yes, I write 17, not 17,000 or something the like. This is a thousandth of the cumulative value of the stock market. The eventual cumulative value of an investment in the stock market seems disconnected from the eventual cumulative value of real economic activity. In the long run, there is an enormous difference between something that grows by 2% per annum and something that grows by almost 7% per annum.

Figure 1 illustrates how the value of an investment in the US stock market has developed over the last 150 years and contrasts it with the development in economic activity, normalized to USD 1 in 1871.

Figure 1. Development in the cumulative value of an investment of one dollar in the US stock market in 1871, inflation-adjusted, and real US GNP per capita, normalized to one in 1871.
Source: From Main Street to Wall Street

In such a graph – where the cumulative value of the stock-market investment dominates so clearly – it almost seems as if there has been no growth in the US economy over the past 150 years. This is not true. Over such a long period, two percent annual growth in economic activity is in fact amazing. A country that grows by two percent per annum over 150 years ends up being one of the richest countries in the world. So, two percent growth in economic activity is a lot when sustained over such a long period. It just pales in comparison with almost 7% growth in the value of a stock market investment.

Another way to illustrate the lack of a simple relation between long-run economic growth and stock returns is to compare the experience of difference countries. Intuitively, one might expect that a country that has grown a lot historically is also a country that has had high stock returns. Again, the data do not back up this intuition.

Figure 2 shows average annual real stock returns and growth rates of real GDP per capita for a number of countries since 1900. The countries are ranked according to the size of average returns.

Figure 2. Real per capita GDP growth and real stock returns for selected countries. 1900–2016.
Source: From Main Street to Wall Street

The US stock market has delivered the highest average annual return, at close to 7%. Has the US economy also been growing the fastest? No. During the past century, the fastest growing economy has been the Japanese. But the Japanese stock market has delivered only half of the return the US stock market has delivered. Similarly, the French economy has been doing fine, but its average annual real stock return has been mediocre. And so on. There is only little evidence that fast growing economies have also delivered higher stock returns.

Thinking deeper about this, it might not be that surprising after all. Stock returns are the sum of a risk-free return and a risk premium. The risk-free rate is in most economic models tied to the growth rate of economic activity. Hence, when we add a risk premium, developments in economic activity and stock returns depart. The question thus becomes what determines the risk premium. I discuss this in the book.

When making comparisons such as those presented here, it becomes relevant whether one uses arithmetic or geometric averages. The volatility of economic growth and stock returns should be discussed too. One can also look at other periods, other measures of economic activity, and so on. I deal with all this in the book.

The relation between stock prices and economic activity

Does the stylized fact that average long-run growth rates of economic activity and average long-run stock returns differ imply that stock markets are living a life of their own, i.e. without any connection to underlying economic activity? The answer to this question is also “no”.

Stock returns are given by stock price gains and dividend yields. It turns out that growth in stock prices (not returns) in the long run is related to long-run growth in economic activity. Many economic models and an abundance of empirical evidence find that stock prices, dividends (not dividend yields), and economic activity share common long-run growth trends. With a fancy word, we say that stock prices and dividends/economic activity are cointegrated. Cointegration means that two variables that grow over time are driven by the same underlying stochastic growth trend, implying that a linear combination of the two series (for instance the difference between the two) does not grow over time. Lots of empirical work has demonstrated that stock prices and dividends are cointegrated, starting with the by-now famous articles by Campbell & Shiller in the late 1980s (link). I have also done extensive research demonstrating cointegration between stock prices and dividends/economic (link).

In my book, I do not conduct statistical tests of cointegration and so on. It is not that kind of book. But I explain what it means and I illustrate it. To provide just one illustration here, Figure 3 shows the development in the ratio of US real stock prices (S&P 500) to US real per capita GDP over the past 150 years, normalized to one in 1871.

Figure 3. The stock-price GNP per capita ratio normalized to 1 in 1871 and its ten-year rolling average. 1871–2018.
Source: From Main Street to Wall Street

The conclusion from Figure 3 is that there is a tendency that the level of US stock prices and the level of US economic activity follow each other in the long run. The ratio of the two is normalized to “1” in 1871. It keeps being attracted to “1” throughout 150 years. There are fluctuations around the mean (“1”), but always a tendency to revert to the mean. The economic interpretation is that there have been (long) periods when the S&P 500 grew faster than economy activity (such as during the pre-WWI period, during the 1950s and 1960s, and during the past four decades), but such periods are followed by (long) periods when the economy grows faster than the stock market, such that the ratio of the two series reverts to its long-run mean. Over 150 years, stock prices follow underlying economic activity.

In the book, I discuss the international evidence, the relation between economic activity and dividends/earnings, and many other interesting topics. The bottom line is that stock returns are not related to the growth rate of economic activity in a simple linear fashion in the long run, as we otherwise might have thought, but stock prices relate to the level of underlying economic activity.

The last four decades

The final topic I would like to discuss briefly today is the fact that the valuation of stocks has soared, according to a number of metrics, during the past four decades.

Figure 3, for instance, shows that the S&P 500 currently trades at a historically high level relative to US GDP. If one relates stock prices to earnings or dividends, today’s stock-market valuation seems even more extreme. I show this in the book. Stocks seem “expensive”.

The final graph today, Figure 4, shows the total value of the US stock market in relation to total US GDP.

Figure 4. The value of the aggregate stock market relative to aggregate GDP.
Source: From Main Street to Wall Street

In popular writing, this is also called the “Buffet Indicator” (link). People use different measures of the value of the stock market when calculating this measure. I use the total value of US corporate equity. As is clear from Figure 4, during the past four decades, growth in the value of US corporate equity has outpaced growth in the US economy. Warren Buffet uses this to argue that future stock returns will be low, as stocks have now become “expensive” in relation to the value of economic activity. Other people (link) use it to argue that the rewards to economic growth have been reallocated from workers to equity owners (helping to explain developments in inequality). In my book, I discuss some of the reasons we have seen an increase in valuations, such as the growing use of share buybacks, low interest rates, and other contributing factors. I also discuss whether it indicates low returns going forward, something that we will also return to on this blog.

Conclusion

This post has discussed some of the topics that I deal with in the initial parts of From Main Street to Wall Street. Those initial parts examine the long-run relation between economic activity and the stock market. I can, though, only provide a few glimpses here. The book discusses many additional topics in relation to long-run growth and returns, such as what is actually economic activity, are people “happier” when economic growth is high, growth and inequality, stock return decompositions, measures of risk, risk aversion, the equity risk premium, and so on and so forth. You have to read the book if you want to know more about these exiting topics.

In my next blog post, I will turn to a few of the book’s conclusions regarding the relation between economic activity and financial markets over the business cycle. 

From Main Street to Wall Street

My book – From Main Street to Wall Street (link) – has been published. This blog post explains why I wrote the book and its contents. The next three posts (that will be sent out in due course) will present some of the analyses and conclusions from the book.

Background

Let me start in 2008. The financial crisis was at its worst.

I was a young professor of finance. I studied financial crises (my Ph.D. is on currency crises) and the relation between the macroeconomy and expected stock returns.

One day, I was invited to a TV program. I was asked to explain something in relation to the crisis. I do not remember the exact topic, but the appearance revealed that I have an ability to explain complicated financial issues in simple terms. There was a need for people with such skills, as many found it difficult to understand what was going on during the financial crisis. I decided that it was part of my duty as a professor at a publicly funded university to help people understand the complicated interactions between the financial sector and the real economy that were brought to surface during the financial crisis. Back then, I became something like an “academic TV star” in Denmark.

A couple of years after the crisis, in 2012, the then Danish government established a commission. Its task was to evaluate the causes and consequences of the financial crisis in Denmark. Due to my presence in the Danish debate and my research on this area, I was asked to be its chairman. I accepted the offer.

The next couple of years, I worked day and night. The report – today called the “Rangvid-report” (link) – was published in autumn 2013.

The decision to write a book

At some point in 2015, or was it 2016, i.e. a couple of years after the report, I started talking to myself about my next big project. With “project financial crisis” completed, I felt I somehow needed a new project. I decided to write a book.

I did not want to stress myself. I have been writing when time has allowed it. Sometimes, I have been able to write a lot. At other times, several months have passed when I was not able to write anything. This has been fine. No stress. In fact, an enjoyable process.

What should its topic be? The obvious topic would have been the financial crisis, given my involvement here. But, as mentioned, I had just completed writing the report of the financial crisis, and everybody else were writing books on the financial crisis. It should be something different.

I am and always have been fascinated by financial economics. Economic growth determines so many things in life. Financial markets help people fulfill some of their big dreams: buy a house, save for retirement. I have done research on the relation between the macroeconomy and financial markets. I decided the book should deal with this.

I am an empirical economist. I have done formal theory when I was young, but I have to come to realize that I do not have a comparative advantage here. Hence, I wanted the book to be a fact-based book on the relation between the economy and financial markets.

I started investigating the market. I was surprised.

Every day, newspapers and business magazines cover the latest macroeconomic news, not least to help investors navigate financial markets. Is falling unemployment good for stocks because it indicates the economy is doing well? Or, is it bad because the central bank will then start tightening monetary policy, potentially hurting stocks? Will countries expected to grow fast provide higher returns? How will economic developments affect financial markets?

Given the attention academics and investors devote to understanding the economy and its impact on financial markets, it was somewhat surprising to me that no well-known book focusing on the relation between the economy and the stock market, and targeted towards a broader audience, existed. I decided that my book should aim at filling this gap.

Of course, there are related books. The one closest to mine is probably “The long good buy” by Peter Oppenheimer (link). I like to think that my book is more thorough and has a firmer anchoring in the academic theories, given my background as an academic and Oppenheimer’s as a practitioner. Also, my book has this structure with the long and the shorter run. Lastly, and probably most importantly, I carefully explain how we can use the insights from the theories and historical evidence to formulate predictions about the future. This being said, Oppenheimer’s book is a great extra reading if you are interested in these topics.

Other related books could be “Stocks for the Long Run” by Jeremy Siegel (link), “Expected Returns: An Investor’s Guide to Harvesting Market Rewards” by Antti Ilmanen (link), and “Financial Markets and the Real Economy”, edited by John Cochrane (link). These are all great books, but they cover different aspects of the topic. Siegel’s book is a description of the stock market on the long run, whereas my book relates the stock markets to the underlying economy, over the short and the long run. “Expected Returns” focuses on providing a comprehensive overview of theories of expected returns on different asset classes, whereas my book focuses on the relation between the macroeconomic and equity markets, over the long run and the business cycle. “Financial Markets and the Real Economy” is a collection of academic articles written by many different authors, i.e. it serves a very different purpose and targets a very different audience than my book, even when the title indicates the same topic.

Writing style

The book is written in a style similar to this blog. Of course, this is a blog, so my language here is sometimes less formal than what I use in the book, but the use of tables and figures to provide a fact-based background to the relation between financial markets and the economy is similar to this blog.

I decided that my book should be academic in nature, but written in a style that is accessible to a broader audience. For instance, the book does not contain formulas or equations, except like “Real equity returns = Nominal equity returns – inflation, “Equity returns = dividend yield + capital gains”, and the like. The book presents the academic theories relating to the different topics, but in a way that should be understandable for interested non-academic readers and students.

An important characteristic of the book is its reliance on the data. Figures and tables richly illustrate and back up arguments and theories. The book is empirical in nature.

I wanted the book to be reasonably comprehensive. Its purpose is to give the reader a thorough understanding of how economic activity affects equity returns. Also, in some chapters, equity returns are compared to interest rates and bond returns.

All chapters end with a checkpoint list that summarizes key insights.

Table of Contents

A guiding principle in the book is that movements in economic activity contain a long-run and a business-cycle component. Over the long run, economies grow. Over shorter horizons – over the business cycle – economic activity fluctuates. My book examines both the long-run relation between economic growth and stock returns as well as the business-cycle relation.

After a first part that introduces fundamental stylized facts and important concepts, the remaining parts of the book are structured around the long-run and short-run relations. In my book, the long run refers to multiple years or decades. The short run to months or a few years.

Over the long run, economies grow. The rate at which an economy grows over the long run is the determinant of whether a country is rich or poor. Expectations to developments in economic activity over the long run help us formulating expectations to returns from financial assets over the long run. When a 25-year old individual asks how much he/she should save to live a decent life after retirement, i.e. after many years, the answer will not least depend on what we expect financial assets to return over the next many years. And, the answer to that question in turn depends on how fast we expect economy activity to grow over the long run. The second part of this book explains what we know about long-run economic growth, about returns on stocks over the long run, and about the relation between long-run economic growth and stock returns.

Over the short run, economic activity fluctuates. Sometimes even substantially so. There are years when the economy is booming, unemployment is low, and times are good. At other times, in recessions, economic activity falls, people are laid off, and times are rough. The recurrent alternations between good times and rough times is the business cycle. Economic activity and the stock market share a common business-cycle pattern. The level of economic activity and the value of stocks rise and fall jointly throughout the business cycle. In order to understand why stock returns are sometimes high and why they are sometimes low, it helps to understand business-cycle fluctuations in economic activity. The third part of the book explains how economic activity and stock returns stocks are related over the business cycle. This part also describes monetary policy, as monetary policy plays an important role in understanding business-cycle fluctuations and stock markets. This part of the book also devotes one chapter to the financial crisis of 2008-2009, as an example of an economic and financial event that dramatically influenced economic activity and equity markets.

When we understand how economic activity relates to financial assets, over the long run and business cycles, we are better equipped to formulate expectations to future returns on financial assets, over the short run and over the long run. This is what the fourth part of the book deals with. It explains how economists judge the outlook for the economy, and it takes us through short-run and long-run stock-return predictability. I emphasize that there is considerable uncertainty surrounding return forecasts, but also that stock returns contain a small degree of predictability. Even if it seems small at first glance, it accumulates and has important implications for academics and practitioners alike.

A final, short, section contains my view on a few practical investment advices.

Publication process

Oxford University Press is the publisher of the book. OUP has done a great job setting up the book. It looks good. The cover is also cool. I am happy.

I handed in the manuscript in spring last year, right before corona arrived. This means that I did not manage to include a chapter on the corona-crisis. If there ever will be a second edition, it will include such a chapter. Until then, many posts on this blog describe different aspects of how the corona-crisis has affected the economy and stock markets.

I received the book in print a couple of weeks ago. It is a nice feeling having published a book.

Conclusion

This post has provided some background on why I wrote From Main Street to Wall Street and its contents. My next three posts will describe some of its key conclusions. The first post will be on the long-run relation between the economy and the stock market, the next on the business-cycle relation, and the final on how we can use those relations to say something about expected future returns.

Yields and inflation expectations

During the last couple of weeks, yields have been rising and stock markets falling. Standard market turbulence is not interesting for this blog – stocks go up and down, most of the time up, and yields fluctuate – but intriguing (and expected) patterns characterize recent events.

Everybody seems to agree what is going on markets these weeks: Vaccines are successful and being rolled out, so economies will open up soon, and Biden will get his stimulus package to the tune of USD1.9tn. These two things (an already strong economy when opening up and on top of that a large stimulus package) will lead to very strong growth during the second half of 2021. Inflation will rise and the Fed will have to tighten monetary policy. The expected rise in inflation and the policy rate leads to increases in yields today. This hurts stocks. These US developments spill over to other countries.

This story largely makes sense. Looking at the data, however, interesting outstanding issues remain.

The story

Inflation expectations have been on the rise since the start of the rally in April 2020. Figure 1 shows developments in expected average annual inflation in the US over the next five years and the yield on five-year Treasuries since April 1, 2020:

Figure 1. Yield on 5-year Treasuries and 5-year breakeven inflation rate. Daily data since April, 2020. Two percent inflation target indicated by dotted line.
Source: Fed St. Louis Database

From a low level, expected inflation rose strongly during the summer-2020 rebound in economic activity, i.e. from April until August, and then stabilized. Since the election of Biden in November, inflation expectations have been on the rise again, because of the expected arrival of vaccines and the stimulus package.

Today, early March 2021, financial markets expect US inflation to be 2.4% per year on average over the next five years. Early November, expected inflation was 1.6%. An increase of almost one percentage point over the course of four months.

An expected rate of inflation of 2.4% is above the Fed’s target rate of inflation of 2% (link). With its new policy, the Fed might allow inflation to exceed 2% for some time, following a period of low inflation, such that average inflation over time approaches 2% (link). But how much more than 2% inflation will the Fed allow before it reacts? At 2.4%, markets start speculating that the Fed will raise rates to keep inflation expectations anchored.

Yields on 5-year Treasuries (also shown in Figure 1) had not moved until a few weeks ago. This also makes sense. As long as inflation is below the Fed’s two percent inflation target and the economy is suffering from the corona-recession, nobody expects the Fed to raise rates. As inflation started exceeding the target, expectations of a Fed hike started to be priced in.

Yields on one- and two-year Treasuries have not started rising yet, whereas yields on Treasuries with a maturity exceeding five years (five-year, seven-year, ten-year, twenty-year, etc.) have increased markedly during recent weeks. This means that the Fed is expected to start raising the policy rate in a couple of years only.

The rise in yields has caused turbulence on stock markets. Volatility (standard deviation of daily changes in the Nasdaq index, for instance) was 1.1% during November and December 2020 but 1.4% during January and February 2021. The same goes for the SP500 (0.83% during the last two months of 2020 vs. 1.04% during the first two months of this year). Stocks suffer.

Complications

All this (yields rise when expected inflation exceeds the target of two percent) seems fine and makes sense. The complicating feature, however, is that ten-year yields (and twenty-year and thirty-year, i.e. yields on long-maturity securities) have been constantly rising in relation to shorter-term yields since markets calmed down in April 2020:

Figure 2. Yield on 10-year Treasuries minus yield on 5-Treasuries. Daily data since April, 2020.
Source: Fed St. Louis Database

In April 2020, 10-year yields were 25 basis points above 5-year yields. Today, they are 80 basis points above.

It is thus not a new thing that yields rise. Longer-term yields have been doing so for some time. The new thing is that other yields start to rise, and that the increase in longer-term yields has accelerated during recent weeks.

We have just argued that it makes sense that markets start speculating that the Fed will raise rates when inflation exceeds two percent, and yields consequently react. But expected inflation over the next ten (and five) years has been below 2% ever since the start of the corona crisis and up until the turn of the year. Nevertheless, yields on ten-year (and longer) securities have been rising since April.

It is instructive to split expected inflation over the next ten years into expected inflation over the next five years and the subsequent five years, i.e. years 1-5 and 6-10 (the latter is sometimes called “5-year, 5-year expected inflation” or “5y5y expected inflation”):

Figure 3. 5-year breakeven inflation rate, 10-year breakeven inflation rate, and 5-year,5-year forward inflation expectation. Daily data since April, 2020. Two percent inflation target indicated by dotted line.
Source: Fed St. Louis Database

Up until January 2021, expected inflation was generally below the 2% target rate of inflation. Shorter-term (1-5 years) expected inflation was lower than long-term expected inflation (6-10 years) but also rose faster. Still, ten-year yields rose while shorter-term yields did not.

Expected inflation over the next ten years is the average of the 5 year and 5y5y expected inflation.

The complication, thus, is that while it is fine and makes perfect sense that market reacts when expected inflation exceeds 2%, it is somewhat puzzling that some yields (longer-term) react when inflation is below 2% while other do not (shorter-term), even when shorter-term inflation expectations rise faster that longer-term expectations?

(Of course, other things, such as the real rate and risk premiums, determined yields, and I briefly discuss these below, but given that there is so much focus on inflation expectations these days, we need to complete that story first).

This issue is a little bit that either you say:

“Given that expected inflation was below 2% until December 2020, it makes sense that shorter-term yields were flat until the start of this year. When expected inflation over the next 5 years started exceeding 2%, 5-year yields started increasing and longer-term yields accelerated”.

Or you say:

“I believe ten-year yields have increased since April because inflation expectations have been increasing, recognizing that expected inflation was below 2%”.

You might be able to come up with a story explaining developments before January, i.e. long-term yields rising but not short-term, and expected inflation below 2%, but it is not straightforward. One story could be that people expected the Fed to keep rates low over the next couple of years, but raise them later. This story would not be based on expected inflation, though, as expected inflation was below 2%, both on the short and the long run.

Before corona

By the way, before we continue, there actually was some relation between expected inflation and yields before the corona crisis, as there generally is. Between autumn 2018 and the corona crisis, expected inflation was falling by close to half a percentage point:

Figure 4. 5-year breakeven inflation rate and 5-year,5-year forward inflation expectation. Daily data since January, 2018. Two percent inflation target indicated by dotted line.
Source: Fed St. Louis Database

And yields were falling, too, during the same period:

Figure 5. Yield on 10-year Treasuries and yield on 5-year Treasuries. Daily data since January, 2018.
Source: Fed St. Louis Database

Real yields

Have real ten-year yields been rising between April 2020 and January 2021, explaining the rise in nominal yields? No. The fact that inflation expectations have been rising faster than nominal yields means that real yields have been falling since April 2020. And, the fact that shorter-term inflation expectations have been rising even faster than longer-term inflation expectations means that short-run real yields have been falling faster than long-run real yields:

Figure 6. Yield on 5-year and 10-year Treasury inflation-indexed security. Daily data since January, 2018.
Source: Fed St. Louis Database

Where should we look?

Given that the relation between expected inflation and yields is tricky (appears to be there now, but was muddy before January), it seems we need to think in terms of risk premiums. This is difficult. Perhaps the inflation risk premium has risen. Perhaps investors before January were more uncertain about inflation after five years than about inflation over the next five years, and demanded a compensation for this. It does not seem the most plausible story, but it is of course a possibility.

The obvious thing to shout is probably “debt”. It seems too early to go down that route, though. I present no analysis here leading credence to this story. And, there are many other potential explanations. It might be debt. But it might also be demographics, productivity, uncertainties, etc. There are many possibilities. The one that faces a hard time is trying to explain all developments in yields since the start of this rally by rising inflation expectations.

Last piece of evidence

The fact that ten-year yields have been rising faster than short-term yields during this recession is not unusual. Historically, around recessions, ten-year yields rise more than short-term yields, either during or right after the end of recessions (sometimes even right before recessions). This is what we have been seeing since April. The slope of the yield curve becomes steeper around recessions:

Figure 7. Yield on 10-year Treasuries minus yield on 5-Treasuries. Monthly data since 1983. Recessions indicated by shading.
Source: Fed St. Louis Database

The point I have been trying to make here is that the increase in the slope of the term structure (long-term yields rising faster than short-term yields) might not be unusual, but it is difficult to explain developments since April by simply saying “inflation is coming”. Expected inflation helps you explain some things, but not all.

Conclusion

Yields are currently going up because inflation expectations have started exceeding the 2% target rate of inflation. People start expecting the Fed to react at some point. This is fine and make sense. But inflation expectations have been going up for longer, since April 2020, and so have long-term yields, even when inflation expectations remained below two percent. Why did some yields (long-term) rise while others did not (short-term) when inflation was expected to remain below two percent and short-term inflation expectations rose faster than long-term expectations?

The global cost of the crisis

Based on IMF forecasts, I calculate the global cost of the crisis as foregone (because of the pandemic) global economic activity up until 2024. The cost amounts to USD 23,600bn, a quarter of global output in 2019. The cumulative output loss is almost twice as large in developing and emerging economies as in advanced economies, though China is an outlier. The calculation represents a lower bound on the final cost, as economic activity might not recover until 2024. Also, the calculation does not include health-induced costs, the inclusion of which would further increase the cost.

As the final (at least for the time being) part of my analyses of the cost of the crisis (link and link), I present here my calculation of the global cost of the crisis.

I calculate the global cost of the crisis as the cumulative loss of global economic activity due to the crisis. This crisis is a health-induced crisis, though. In my previous posts, I calculated both the loss of economic activity and health-induced costs for Denmark. I am not aware of methods to estimate the size of health-induced costs on a global level. Hence, I restrict this analysis to the loss of global economic activity, but discuss health-induced costs.

Expected growth rates

I base my calculations on IMF forecasts for global economic activity. I take IMF’s forecasts from October 2019, when nobody expected the arrival of the pandemic, and compare them to the recent January 2021 IMF forecasts for global economic activity. Figure 1 summarizes IMF’s expectations:

Figure 1. Growth in global output during 2020 and the average annual growth rate during 2021-24. IMF forecasts from October 2019 and January 2021.
Source: IMF WEOs.

In October 2019, IMF expected global output to increase by 3.4% in real terms during 2020. There was no talk about a pandemic. Instead, the major risk mentioned was the uncertainty surrounding global trade and geopolitics (Trump and China).

2020 turned out to be so different. In their recent January 2021 WEO update, IMF now expects global output to contract by 3.5% in 2020. This means that the forecast error (= expectation in October 2019 – realized growth in 2020) amounts to 7%-point.

The pandemic and the resulting lockdowns caused this unprecedented recession. As a comparison, global output contracted by 0.1% “only” because of the financial crisis in 2009. In 2020, as mentioned, the contraction amounts to 3.5%. Globally, this recession has been so much more severe than the one following the financial crisis in 2009.

We do not have the final figures for 2020 yet. The 3.5% contraction for 2020 is a very good guess, though, given that we have the figures for the first three quarters of 2020 and higher frequency indicators for the last quarter, such as industrial production etc. Hence, I would be surprised if the final figure for 2020 turns out to be much different from this number, i.e. from a 3.5% reduction in global output.

Figure 1 also shows that global growth expectations for 2021-2024 have been revised upwards, compared to what was expected before the pandemic. In October 2019, IMF expected global output to increase by 3.6% per year on average during 2021-24. In January 2021, IMF expects the world economy to grow by 4.2% per annum during 2021-24. So, 2020 was worse than expected before the pandemic but 2021-24 is expected to be better. What do the revisions imply for the expected cumulative loss of economic activity during the 2020-2024 period?

The global loss of economic activity

Before the pandemic, in 2019, aggregate global output amounted to app. USD 87,500bn. In this figure, I calculate the path of global output expected before the pandemic (using growth rate expectations from October 2019) and contrast it with the path expected now, i.e. taking into account the actual 2020 development and current (January 2021) forecasts for global growth during 2021-24:

Figure 2. Paths of real global output, based on IMF forecasts from October 2019 and January 2021. Billions USD.
Source: IMF WEOs and own calculations

In October 2019, the expectation was that global output would amount to app. USD 90,500bn in 2020 (in real terms). Instead, global output amounts to app. USD 84,500bn in 2020. In 2020, the world earned app. USD 6,000bn less because of the pandemic.

Figure 2 reveals that global output will most likely remain below the level expected before the pandemic during the next couple of years, in spite of the increase in expected growth over the next four years that Figure 1 illustrated.

In aggregate, during the 2020-2024 period, the difference between what was expected before the pandemic and what was realized in 2020 plus what is expected going forward from here amounts to USD 23,600bn. This is my estimate of the global cost of the pandemic. It is the value of foregone economic activity in 2020 plus the expected value of foregone economic activity up until 2024.

USD 23,600bn is a big number. To illustrate, it is 27% of 2019 global output. A quarter of one year’s global output has been lost due to the pandemic. It also (more or less) corresponds to the value of everything that is produced (GDP) in the still-largest economy in the world, the US, during one year. Or, more or less, one and a half times everything produced in China in a year. As a final illustration, it is app. 12 times Biden’s new stimulus package of USD 1,900bn.

Advanced vs. Developing and Emerging economies

The pandemic is global. The cost of the pandemic will not be shared equally among Advanced and Developing/Emerging economies, though.

This figure shows that the forecast error for 2020, i.e. the difference between expected 2020 growth and actual 2020 growth, was equally large for Advanced and Developing/Emerging economies. For both Advanced and Developing/Emerging economies, growth during 2020 turned out to be 6%-7%-points lower than expected before the pandemic:

Figure 3. Forecasts error for 2020 and forecast revisions for 2021-24 for output growth in Advanced and Developing/Emerging economies.
Source: IMF WEOs.

Advanced economies have provided substantial fiscal and monetary support to households and firms in 2020, as described elsewhere on this blog, helping to contain the contraction. Also, vaccines are expected to be widely available in Advanced economies during 2021, supporting the recovery as of summer 2021.

In contrast, it will take longer for vaccines to be rolled out in Developing/Emerging economies. Also, oil exporting and tourism-dependent Developing/Emerging economies have suffered particularly much, IMF mentions.

China differs from most other Developing/Emerging economies, as China has been successful in containing the pandemic after the initial outbreak in early 2020. Also, China has provided substantial fiscal and monetary support.

In total, this means that the cost of the pandemic will not be shared equally around the world. The output loss due to the pandemic will be 17% of 2019 GDP for Advanced economies, but will be almost twice as large, 31%, in Developing and Emerging economies. In Chinas, it will only be 7%:

The cost of the crisis in different parts of the world. Costs are the reductions in economic activity from 2020-2024 because of the pandemic, in relation to output in 2019.
Source: IMF WEOs and own calculations.

Uncertainty

There is considerable uncertainty surrounding the recovery. Will vaccines be rolled out according to plan, will we face new mutations that vaccines do not protect against, will consumers increase spending as much as we expect when restrictions as removed, etc.?

IMF presents up- and a downside forecasts that reflect this uncertainty. The calculations above are reasonably robust to the future risk scenarios. The reason is that the cost is mainly due to the contraction in 2020, and there is little uncertainty surrounding that by now.

Figure 4 presents the path of world output as expected before the pandemic and how it is expected to develop as of now, including up- and down-side scenarios:

Figure 4. Paths of real global output, based on IMF forecasts from October 2019 and January 2021. The January 2021 forecasts include an upside and a downside scenario. Billions USD.
Source: IMF WEOs and own calculations.

The magnitude of the cost of the crisis is not much affected by this uncertainty. The cost of the crisis is reduced to 24% (from 27%) of 2019 global output in the positive scenario but increases to 29% of 2019 global output in the downside scenario.

Health-induced costs

This analysis calculates the cost of the crisis in terms of foregone output around the world. The crisis is a health-induced crisis, though. In my previous posts that estimated the costs of the crisis in Denmark , I factored in the economic value of the costs of premature mortality, health impairments, and mental health impairments.

A similar global calculation requires estimates of the value of a global statistical life, as well as estimates of the number of people experiencing severe health impairments because of the virus, as well as the number of people facing mental health impairments. I know of no good way to calculate these such on a global level. Hence, in this post, I calculate the part of the cost due to foregone economic activity. This is certainly an important number to know in itself.

To give one perspective on the health-induced costs, though, Cutler & Summers (link) estimate that the economic value of health-induced costs in the US is, largely, equal to the loss of economic activity for the US. They estimate the total US loss at USD 17 trillion and the loss of economic activity in the US to USD 7.5 trillion, i.e. the loss of US economic activity corresponds to app. 50% of the total US loss. I estimate that health-induced costs are significant for Denmark, too (link and link), but relatively smaller than those calculated by Cutler & Summers for the US.

Conclusion

The global recession caused by the pandemic has been unprecedented in terms of both its magnitude, its cause, as well as the policy response. Based on IMF forecasts for global economic activity up until 2024, I calculate the global cost of the crisis in terms of foregone economic activity around the world. The global cost of the crisis amounts to app. USD 23,600 bn. It is an unfathomable number. It is the value of the reduction in cumulative global output resulting from the crisis. It is income that would have been generated around the world was it not for the pandemic. It corresponds to a quarter of total economic output in 2019.

As I do not have good estimates of global health-induced costs, I restrict the analysis to foregone economic activity. On the one hand, this means that the analysis is not an estimation of the total cost of the crisis (loss of economic activity + health-induced costs). On the other hand, it is certainly important to know the value of the loss of economic activity in itself.

I base my calculation on IMF data. In October 2019, IMF presented forecasts up until 2024. Hence, my calculations run until 2024. As shown in some of the figures above, global economic activity will in 2024 most likely not have reached the level expected before the crisis. Hence, the eventual global cost of the crisis might end out being higher than the number presented in this analysis.

The cost of the second wave

Early December 2020, I presented my calculation of the expected cost of the corona crisis in Denmark, taking into account both economic and health-related costs (link). Since then, the situation has turned to the worse, and the expected cost of the crisis has increased by something like 50%. The calculation here is done for Denmark, but given similar types of waves in the US, the UK, and many other countries in Europa, I would expect similar types of consequences.

This graph shows the sad development in the number of people dying with corona in Denmark since March 11, 2020, on a daily basis.

Daily Covid-19 associated deaths in Denmark. Blue columns: March 11, 2020 through November 26, 2020. Red columns: November 27, 2020 through January 6, 2021.
Source: https://en.ssi.dk/.

The blue columns show the daily number of deaths up until November 26, 2020, when I did my initial calculation. The red columns show the numbers since then. From March through November, i.e. including the summer months when few people got infected, approximately 100 people died with corona per month. During the past six weeks, approximately 100 people died per week. During December, more people died with corona than during April.

Economic activity suffers

As a consequence of the dire situation, tough restrictions have been introduced: Max. five people are allowed to meet, people are working from home, “everything” (bars, restaurants, schools, universities, most retail shops, etc.) is closed, etc. Similar restrictions have been imtroduced in many other countries.

Because of the restrictions, economic activity suffers. Forecasts for economic activity have been revised.

In my calculation from early December, I used forecasts from the independent Danish Economic Councils (link). I used their “main scenario” from their autumn 2020 report. The Councils also published a negative scenario. Many economists, including the Council itself, and including myself by the way, now expect that this is a much more likely path for economic activity going forward.

Danish GDP forecasts.
Source: Danish Economic Councils

In their “main scenario”, the Councils expected GDP to fall by 3.6% in 2020, strongly rebound by 3.8% in 2021, and grow by 2.3% per year on average during 2021-2025. Now, a more likely path for Danish GDP is that it falls by 4.2% in 2020, grows by 1.5% in 2021 and by 2.5% per year on average during 2021-2025. In other words, due to the worsening of the situation since late November, the fall in economic activity during 2020 will probably be larger and, in particular, the rebound during 2021 will be smaller and delayed. Given that we start out 2021 in lockdowns, it is more reasonable to expect something like 1.5% growth in 2021, instead of 3.8%.

This figure collects the expected path of GDP without the coronacrisis (“Structural GDP w.o. corona”), the path expected in autumn, and the path that I and many other economists expect now (“Negative scenario”)

Danish GDP forecasts. DKK billion. Main and negative scenarios from Autumn 2020 projections vs. expected path of structural Danish GDP assuming no corona crisis.
Source: Danish Economic Councils

As is obvious from the graph, the expected reduction in GDP is just so much bigger now. More precisely, the expected loss in GDP due to corona (the reduction in GDP due to corona) is now app. DKK 400bn. This is basically two times the loss of economic activity expected just one month ago. This illustrates the costs of an additional wave of the coronacrisis.

Health-induced costs

Health-induced costs depend on the number of deaths, as described in my December blog post (link). In my December post, I assumed that the rate of deaths prevailing at the time of writing would continue for one more year. This was an assumption similar to the one used by Cutler & Summers for the US. This accumulated to 2100 expected deaths in total in Denmark. I argued that this was an aggressive assumption, given that I hoped vaccines were about to be rolled out, but it was the one used by Cutler & Summers, so I used that, too, such that I could compare costs for the US and Denmark.

Now, unfortunately, the assumption does not look that far off. During the past six weeks only, the number of deaths has increased by 75% in Denmark, from 800 to 1400. Given the severity of the current wave, and the slow roll-out of the vaccines, I do not feel comfortable changing my assumption with respect to the health-induced costs.

Updated expected cost of the crisis

This table collects my updated expected economic costs of the corona crisis in Denmark, together with my December expectation:

My estimate of the expected total cost of the crisis has increased from DKK 336bn to 536bn. This corresponds to 25% of pre-crisis Danish GDP, up from 16%. It also corresponds to app. DKK 90,000 per Dane (app. USD 15,000), compared with my best guess of DKK 60,000 per Dane in late November/early December (app. USD 10,000). In just six weeks, due to a terrible second wave of the corona crisis, the cost of the crisis has increased by something like 50% on a per-capita basis.

Conclusion

Obviously, there is a lot of uncertainty surrounding calculations such as these. The mere fact that six weeks of new data can change the economic outlook so much bear this out. Also, I use forecasts from the independent Danish Economic Councils. I expect all forecasters to agree that the economic outlook has turned to the worse, but exactly how much probably differs from forecaster to forecaster.

It is good with one number, DKK 536bn, as we then have something to remember and compare with. With so much uncertainty, on the other hand, we should not be surprised if the final cost will differ from this forecast. In general, whether we talk about forecasts of costs of crises, general economic forecasts of GDP, interest rates, exchange rates, or whatever, it is seldom that forecasts are spot on.

Forecasts such as these help us, though, to understand the magnitude of the crisis. The cost of the crisis (in Denmark) will be in the hundreds of billion Danish kroner. Of this, I am confident. Right now, I believe that something like DKK 500bn, or 25% of GDP, is a good estimate. This estimate might change, but I am reasonably confident that the cost of the crisis will end up being measured in double-digit percentages of GDP. This is the important take-away, not the exact number.

Given these high costs, we should do everything possible to try to reduce them. In particular, the faster vaccines are rolled out, the better. What is difficult to comprehend, I think, is that measures to expand vaccine production capacities were not taken in autumn, when everybody could see that the vaccines were on their way (e.g., by licensing production of vaccines to factories with spare production capacity, etc.). I thus cross my fingers that the rate of vaccination will increase significantly within a not too distant future, thereby helping to contain the otherwise very high cost of this crisis. A Danish newspaper had this cartoon yesterday, link, showing me arguing for the importance of a fast rollout of the vaccines.

2020

It’s this time of the year. This post recalls events of 2020. It has been such an unusual year, so different from what we expected. Luckily, there seems to be light at the end of the very long and dark tunnel, and – I hope – that 2021 will be considerably more joyful than 2020.

2020 started out so well. The roaring twenties and all that. Wuhan was a city I had not heard of, corona a beer people tell me is best served ice-cold with a slice of lime (I do not drink beer, tough I do enjoy wine), and social distancing words we would only get to know too well. Today, we know that Wuhan is a Chinese city with more than eleven million inhabitants and a marketplace where it presumably all started, corona also means something terrible, and social interaction is an activity we have come to miss so dearly.

At the time of writing, app. 75 million cases of corona/COVID-19/SARS-CoV-2 have been confirmed globally and app. 1.7 million people have passed away because of corona. Most countries have been in lockdowns, many still are (again), and the social and economic costs of the crisis have been enormous.

I started this blog in April 2020. This had nothing to do with corona. I had wanted to set up a blog for some years (people ask me where I find time for this, and I really do not know, but seemingly I simply like writing economic stories and analyses). Starting the blog in April this year, however, naturally implied that many of the blog posts have dealt with various economic and financial aspects of the pandemic. In this post, I will review some of the learnings from 2020.

The worst recession on record. With the highest growth rate on record

The recession started in February 2020 in, e.g., the US. Initially, it was caused by a supply shock: lockdowns were imposed and firms could not sell their goods and services and households could not go shopping. In April, when the IMF released their Spring Outlook, they labelled it “The Great Lockdown”. This was a suitable label. The IMF also noted that “This is a crisis like no other” and that “many countries now face multiple crises—a health crisis, a financial crisis, and a collapse in commodity prices, which interact in complex ways”. As unemployment and bankruptcies increased, households and firm got nervous, and demand suffered, too.

The path of economic activity has been highly unusual. This graph shows the quarterly percentage changes in US real GDP since 1947:

Quarterly percentage changes in US real Gross Domestic Product. 2020 encircled.
Source: Fed St. Louis Database

2020 is very much an outlier. On average, from 1947 through 2020, real GDP has grown by 0.8% per quarter. Until 2020, quarterly growth had never exceeded 4%. Economic activity had never contracted by more than 2.6%. Then came the Great Lockdown. During the second quarter of this year, economic activity contracted by 9%. This is almost four times more than the otherwise worst contraction on record. In this sense, it was the worst recession ever.

It has also been the weirdest recession ever. During this recession, we have also witnessed the highest growth rate on record: economic activity expanded by 7.4% during Q3. This is twice as much as the otherwise highest growth rate on record.

This puzzling feature of the recession led me wondering what a recession really is (link). I expressed sympathy with members of the NBER Recession Dating Committee. They face a particularly difficult task this year. Should they conclude that we had one V in spring, with the recession ending in late April, and then a new V now, i.e. two separate Vs (VV), or that we have had one long recession with a double dip, i.e. a double-V (W)? Does it make sense to call it a recession when we experienced the fastest rate of growth in economic activity on record? If you conclude that the economy cannot be in recession when it expands at its fastest growth rate ever, then you must conclude that the recession ended during Q2. But, the NBER Recession Dating Committee has not called the end of the recession yet, i.e., officially, the recession is still ongoing.

You may ask why it is important to know whether the recession ended in April or whether it is still ongoing. The development in economic activity is what it is, whether we call it recession or not. It is important because a “recession” is such an important concept in economics. We inform the public, business leaders, students, and others about the characteristics and consequences of recessions. If a recession can contain the by-far strongest expansion of economic activity on record, we need to change our understanding of recessions.

The very unusual behavior of economic activity during Q2 and Q3 caused very unusual, and scary, developments in unemployment and related aspects of economic activity. This graph shows the monthly change in the number of unemployed in the US:

Monthly changes in the number of unemployed in the US. Millions. 2020 encircled.
Source: Fed St. Louis Database

During March, unemployment in the US increased by 16 million. Again, this was beyond comparison. Until March 2020, the number of unemployed had never increased by more than one million over one month. In March 2020, it increased by 16 million.

As the virus contracted during summer, unemployment fell. There has never been as fast a reduction in the number of unemployed as the one occurring during this summer. In May, the number of unemployed dropped by more than three million. Until May 2020, the number of unemployed had never fallen by more than one million over one month. In May 2020, it fell by more than 3 million. So, within a year, we have had the strongest-ever increase in unemployment, but also the largest-ever fall in unemployment. By far.

Such dramatic events happened all around the world. I documented this here (link) and here (link).

Inspired by these events, I did something admittedly nerdy. I calculated the probability that we would experience events such as these, given the historical data (link). I found the unconditional likelihood that we could see the increase in newly registered unemployed that we saw in spring to be 0.97 x 10^(-841). This is a zero followed by 841 zeroes and then 97. For all practical purposes, this is a zero-probability event. But it did happen. It was just very, very unusual.

The stock market

I use some of my time (a significant part, by the way) to try to understand the stock market. This has not been a straightforward task this year.

Today, the global stock market is 13% percent above its January 1 value, the US stock market is 18% higher, and the Danish stock market 29% higher (MSCI country indices). Given that we have been through the worst recession ever, and that the recession is not officially over yet, this is not what one would have expected prior to the events.

Then, on the other hand, in hindsight it is perhaps not so strange. The recession has been the worst on record, yes. But, we have also had the fastest growth in economic activity on record. I argued (link) that if we imagine that the recession ended in late April, when economic activity bottomed out, the behavior of the stock market fits perfectly well with the historical evidence on the behavior of the stock market.

Central banks have certainly played their role, too. When markets melted down in March, central banks intervened heavily. In contrast to the financial crisis of 2008, it was not banks that were in trouble this time, but firms. Firms could not sell their goods and services due to the lockdown. The limitless purchases of government bonds that central banks have become used to during and after crises thus probably did not do much good (evidence came out that central bank purchases of government bonds are less effective than we are often told, link ). What turned things around, instead, was the announcement on March 23 that the Fed would facilitate credit to firms (link). This was a new policy tool. It led to a complete turnaround of events. I produced this graph (I still think it is a supercool graph): 

Difference between yields on ICE BofA AAA US Corporate Index and 3-month Treasuries (Left hand axis) as well as the SP500 inverted (Right hand axis). Both series normalized to one on January 2, 2020. Vertical line indicates March 23.
Source: Fed St. Louis Database

The graph shows how the stock market lined up with credit spreads. Firms were suffering, and their credit spreads started widening, in late February. The stock market suffered. The Fed announced it would provide credit to firms on March 23. Credit spreads tightened. The stock market cheered. The graph summarizes how the Fed saved credit and equity markets. And, strikingly, the Fed did so by merely announcing they would intervene. Up until today, the Fed has not intervened a lot. In this sense, it was a “Whatever it takes moment of the Fed”.

It should be mentioned that the Fed announced other initiatives on March 23, too, such as the Main Street Lending Program (link) and the Term Asset-Backed Securities Loan Facility (link). The Corporate Credit Facilities were the ones that directly targeted corporate bonds, though. Due to the nature of this crisis, the stock market lined up with credit spreads during this crisis, as the above graph reveals, emphasizing the importance of the announcement of the Corporate Credit Facilities.

Eurozone troubles, or rather no Eurozone troubles

The fact that we have not had to talk a lot about the risk of a Eurozone breakup since summer has been a positive surprise. In spring, there was talk about the risk of a Eurozone crisis. Like so often before, Italian sovereign yields rose relative to German sovereign yields. There was reason to be anxious. I argued that “Some kind of political solution at the EU level would be needed” (link).

This we got. The European Union agreed on a “Recovery and Resilience Facility” (link) that includes both loans and grants. EU has moved one inch closer towards a common fiscal policy. Who will pay is not clear, but EU has shown solidarity. I believe this is positive. At the same time, the European Central Bank continued its interventions and bought a lot of Italian debt. This has kept yields on sovereign bonds low. Here is the Italian-German yield spread during 2020:

Italian yield spread towards Germany. Ten-year government bonds. Daily data: January 2, 2020 – December 21, 2020.
Data source: Thomson Reuter Datastream via Eikon.

Italian yields have been falling continuously since summer, when the EU agreed on its recovery plan. It is positive that we have not had to discuss Eurozone troubles. We have had so many other troubles. Whether this means that we do not have to discuss Eurozone troubles again at some point, I am less sure. But, that is for another day.

Banks have been doing OK

The risk of a Eurozone breakup did not materialize. Another risk that did not materialize was the risk of systemic bank failures. This is positive as well, as economic activity suffers so much more when banks run into trouble and credit consequently does not flow to its productive uses.

During the worst days in March, stress in the banking system intensified. For instance, the spread on unsecure interbank lending increased relative to secure lending:

The TED spread. Difference between 3-Month USD LIBOR and 3-Month US Treasury Bills. Daily data: January 2, 2000 – December 14, 2020. 2020 encircled.
Data source: Fed St. Louis Database.

Stresses lasted only a few days, though. During the financial crisis in 2008, on the other hand, spreads remained elevated for much longer. This time, trust in the banking system was quickly reestablished.

I think I am allowed to claim that this was one of the predictions I got reasonably right. In autumn 2019, when nobody knew about the upcoming crisis, I wrote a policy paper on the Nordic financial sector. It was presented in December 2019 and finally published in June this year (link). I argued that banks are safer today, compared to 2008. Some doubted my conclusion and said, “just wait until the next crisis, then you will see that banks are not safer today”. Well, few months later we had the worst recession ever. Luckily, though, we have not had bank-rescue packages and we have not had to bail out banks. Banks have been withering the storm. In some instances, banks have even been part of the solution by showing flexibility towards troubled firms. I am not saying everything is perfect, but I am saying that the situation has been very different from the situation in 2008. On a personal note, this made me happy, too, as it would have been somewhat embarrassing if banks had failed at the same time I published an analysis arguing that the banking sector is safer. This, luckily, did not happen. Instead, the banking system turned out to be far more resilient than in 2008, as I predicted.

You may add that the Fed rescued markets during spring, as mentioned above, and thereby rescued firms and subsequently banks. True, but there was certainly also tons of rescue packages in autumn 2008. Banks nevertheless failed in large numbers in 2008. They did not this time around. Perhaps, thus, we did learn something from the financial crisis of 2008, and have gotten some things right. This would be no small achievement.

US election and Brexit

There have been other events, for instance the US election and Brexit negotiations. In normal years, such events would potentially have been among the most important events for markets and the economy. This year, the pandemic has certainly been more important. I did manage to write a post on the US election and the stock market, though (link). I discussed evidence that stock markets perform better under Democratic presidents. Only time will tell whether the same will happen under Biden.  

I did not manage to find space to discuss Brexit, but we got a trade agreement on Dec. 24 (link). Hopefully, the EU and UK can now move on.

The cost of the crisis

It is impossible to summarize the pandemic in one number or one word. Hence, I will not attempt to do so. But, I did present a calculation of the expected cost of the crisis in Denmark (link). I arrived at DKK 336bn, or app. USD 10,000 per Dane. This calculation generated some attention in Denmark.

One can discuss every single assumption one needs to make when calculating the expected cost of a crisis: What is the value of a statistical life? What is the value of a statistical life of those who pass away due to COVID-19, i.e. who are typically above 80? What is the past loss as well as the expected future loss in economic activity due to the crisis? Does it make sense to present one number when there is so much uncertainty? And so on. These are all fair points, but if we want to have a meaningful discussion of the impact of the crisis, we have to start somewhere.

In my calculation, I closely followed the assumptions of Cutler & Summers, such that US numbers and Danish numbers can be compared. This allowed me, for instance, to conclude that the cost of the crisis in Denmark, most likely, will be much lower than the cost of the crisis in the US.

Conclusion

I must admit I find it difficult to end this last post of 2020 on a happy note. Right now, at the time of writing, the situation is bad in the country I live, Denmark, and in many other countries in Europe and around the world. Numbers of new cases and deaths have been rising recently, or are on the rise again, and more and more restrictions and lockdowns are being imposed. Days are grey and short. The crisis has already been tremendously costly and it is clearly not over yet.

Nevertheless, I will try to end the post on a positive note. It gives me hope that several countries have started vaccinating people, and it seems to be working well. Finally, the EU also starts vaccinating people now. This has taken way too long, however, given the severity of the crisis and the fact that other countries started weeks ago. And, yes, every day counts. If it is correct, though, and I deliberately write if, that the EU has failed when it comes to the approval process and purchase of vaccines, as the normally well-informed and serious magazine Der Spiegel claims (link), it is a scandal. Biden aims to vaccinate 100m Americans within his first 100 days in office (link), close to a third of the US population. As things look now, it seems unlikely that we will be able to achieve the same in Europe. Christmas is all around us, though, so let us hope that somehow things will develop in the right direction.

Therefore, let me focus on the bright side. With the jabs, the situation will most likely start to improve within a not too distant future. I will try to convince myself that I see weak light at the end of the long and dark tunnel, even when we probably have to wait many months before things really calm down. Days are at least getting longer. I will focus on this, then.

With this, which is meant to be a positive message, let me thank you all for reading this blog and for sending me many encouraging mails with feedback. Please keep on doing so – it is highly appreciated.

I conclude by expressing hope that next year will be considerably more joyful than the one we leave behind.

Happy New Year!

The cost of this crisis

The corona crisis has caused a loss of economic activity. To calculate the total economic cost of this health-induced crisis, we must factor in health-related costs. Based on calculations for the US by Cutler & Summers, this post calculates the economic cost of the crisis in Denmark. Many uncertainties surround such calculations, and I discuss those. It seems a robust conclusion, though, that the cost of the crisis in Denmark will be considerably smaller than the cost of the crisis in the US. The broader implication of this result is that there is considerable variation across countries in the economic cost of the pandemic.

Economic activity suffered dramatically during spring, and it will take time before economic activity has recovered to its pre-crisis growth trend. In addition to the loss of economic activity, we must take into account health-related costs, if we want to estimate the total economic cost of the crisis. This is no easy task. Inevitably, it will be based on a number of assumptions. In a recent Journal of the American Medical Association article, Harvard University Professors David M. Cutler and Lawrence H. Summers (link, link) present a calculation of the total economic cost of the corona crisis in the US. Their calculation takes into account both the cost of lost economic activity as well as the economic value of premature mortality, the economic value of health impairments, and the economic value of mental health impairments.

Cutler & Summers (CS) find huge economic costs of the corona crisis in the US. They estimate the economic cost at USD 16 trillion. This is 90% of US GDP. This exceeds the cost of the financial crisis in the US by a wide margin, Cutler & Summers argue.

This post presents the first calculations of the cost of the crisis in Denmark. I follow the procedure of Cutler & Summers, but use Danish data. Knowing results from other countries, we will be able to better judge whether results for the US are globally representative.

I stress that these are first calculations and that uncertainties surround them. But, if we want to have a meaningful discussion of the cost of this crisis, we have to make assumptions, and then discuss their robustness. I do so.

Lost GDP

The first component of the calculation of the economic cost of the crisis is loss of economic activity. CS look at forecasts for US Gross Domestic Product from the Congressional Budget Office right before the crisis and compare it with the latest forecasts. The difference is the loss of economic activity. CS calculate a loss of USD 7,600bn. This corresponds to 35% of US pre-crisis GDP.

In Denmark, the Danish Economic Councils publish independent forecasts for Danish GDP. In their latest autumn 2020 report, the Councils publish figures for the expected development in structural Danish GDP, had there been no corona crisis. They also estimate expected GDP as of now. The results are here:

Danish GDP forecasts. DKK billions. Autumn 2020 forecast (“GDP”) vs. expected path of structural Danish GDP assuming no corona crisis.
Source: Danish Economic Councils

The difference between now-expected GDP developments and expected GDP developments without the corona crisis is the loss of Danish economic activity due to the crisis. This amounts to DKK 214bn (app. USD 40bn) for the years between 2020 and 2024, when Danish GDP is assumed to have recovered to its without-corona crisis structural trend. This loss of economic activity corresponds to approximately ten percent of Danish 2019 GDP.

One reason why the loss of economic activity in Denmark is expected to be considerably smaller than the loss in the US (10% in Denmark vs. 35% for the US) is that the Congressional Budget Office (CBO) expects the loss of economic output to persist in the US. CS reproduce this figure from the CBO:

US GDP forecasts. July 2020 forecast vs. January 2020 forecast.
Source: Cutler & Summers (2020).

Even ten years out, in 2030, US GDP is expected to be lower than its pre-crisis expected growth path. This is different in Denmark. In Denmark, the most recent forecast indicates that economic activity will have recovered in 2024. This reduces the loss of GDP in Denmark, relative to the US.

Cost of premature mortality

Deaths add to the economic cost of the crisis. CS note that 190,000 Americans had passed away due to Covid-19 by late September 2020. This is 0.06% of the US population.

In late September, when CS wrote their report, 5,000 Americans passed away per week as a result of Covid-19. CS expect this to continue for another year, i.e. an additional 260,000 deaths. CS estimate excess non-Covid-19 deaths at 40% of Covid-19 deaths. This means 1.4 x 450,000 = 625,000 American deaths.

In the US, a statistical life is estimated at USD 10m. CS reduce this by 30% to be conservative. All in all, this results in an economic costs of USD 4.4 trillion due to premature mortality. This is 20% of US GDP.

In Denmark, 816 have passed away as a result of COVID-19, at the time of writing this post (early December). In late September (to compare with the CS US figures), 650 had passed away. This is 0.01% of the Danish population.

This figure shows daily deaths due to Covid-19 in Denmark since March 2020:

Daily Covid-19 associated deaths in Denmark.
Source: https://en.ssi.dk/.

During the last couple of weeks, around 25 Danes have passed away because of Covid-19. Assuming, like CS, that this will continue for another year, this accumulates to 1,300 additional deaths. Assuming 40% excess non-Covid-19 deaths, like CS, results in 2,900 deaths over the next year.

In Denmark, a statistical life is estimated at DKK 34m (link). This corresponds to app. USD 5m. In other words, the value of a statistical life in Denmark is only half the value of a statistical life of an American. To be conservative, like CS, I use 70% of this. This gives an economic cost resulting from premature deaths of DKK 79bn. This is around 4% of Danish GDP. I.e., again, the loss in Denmark seems to be considerably smaller than the comparable loss in the US (4% of GDP vs. 20% of GDP). This is because fewer Danes are expected to pass away because of Covid-19 and because the value of a statistical life is assumed to be considerably smaller in Denmark.

Health impairments

Some of those surviving Covid-19 will face significant long-term health complications. CS mention that there are approximately 7 times as many survivors from severe or critical Covid-19 diseases as there are Covid-19 deaths, and that a third of these will experience long-term complications. CS assume that the cost of this is 35% of a statistical life and that it lasts for one year. This adds USD 2.6 trillion, or 12% of US GDP.

In Denmark, using the procedure of CS, expect 2,100 deaths, as mentioned above. This means 4,900 individuals with long-term health impairments over the next year. Assuming that the complications last for one year, and assuming an economic cost of complications of 35% of a statistical life, this means an economic loss of DKK 41bn, or 2% of GDP. Notice, like CS, I use the conservative value of 70% of a statistical life, i.e. 35% of 0.7 x 34m.

Mental health impairments

Many people get anxious or fell depressed during the pandemic. This could be because people need to isolate at home, and thus face loneliness, because of fear of losing your job, i.e. economic insecurity, fear of contracting the virus, etc. CS report that 40% of American adults have reported symptoms of depression or anxiety during the corona crisis. Normally, CS report, 11% of Americans report these symptoms. CS report that previous studies evaluate the one-year cost of depression and anxiousness at USD 20,000 per case. This amounts to an additional cost of USD 1.6 trillion, or 7.5% of US GDP, for the corona crisis.

In Denmark, we do not have official data on the number of people feeling anxious or depressed during the corona recession. What we do have, on the other hand, is the comprehensive health report of the Danish population from 2015 that also estimates the economic costs of a number of diseases in Denmark (link).

This report mentions that 136,000 Danes suffer from anxiousness and 91,000 from depression (in 2012). In total, around 4% of the Danish population. The report calculates an aggregate cost of lost economic activity of DKK 8.6bn for anxiousness and DKK 3bn for depression, primarily as a result of early retirements. These costs relate to 2012. Since then, inflation has been 7% in Denmark, i.e. the 2020 value is DKK 12.4bn (app. USD 2bn), assuming the same number of depressed and anxious individuals. CS assume an almost fourfold increase in the number of individuals facing depression and anxiousness because of the corona crisis. Given that this number is associated with a large degree of uncertainty in Denmark, and because one might hope that many people will recover, as this is a temporary crisis, such that not all of them will enter into early retirement, I double this number (in contrast to CS who assume a fourfold increase). I.e., I assume that the economic cost of mental health impairment amounts to DKK 24bn (app. USD 4bn), or app. 1% of Danish GDP.

Total costs

In total, taking into account the loss of economic output as well as health costs, reflecting both premature deaths, long-term health impairments, and mental health impairments, the economic cost of the pandemic in Denmark amounts to DKK 336bn, or 16% of Danish GDP:

This corresponds to approximately DKK 60,000 per Dane (app. USD 10,000). For a family of four, this means that the economic cost of the crisis is app. DKK 240,000 (app. USD 40,000).

Implications

A number of implications follow from these calculations:

  • A cost to society of 16% of GDP is an enormous cost.
  • With this in mind – that a 16%-GDP cost is enormous – it also follows immediately, on the other hand, that the cost of the pandemic in Denmark is considerably smaller than the cost of the pandemic in the US. CS find, as mentioned, that the cost of the pandemic in the US will amount to 90% of US GDP. For Denmark, I find it will be 16% of Danish GDP. This implies that there is considerable variation across countries in the expected cost of the pandemic. Or, in other words, that it would be a mistake to assume that the cost of 90% GDP found for the US is a globally representative figure.
  • There are several reasons why the cost of the pandemic is considerably lower in Denmark:
    • Danish GDP is expected to recover faster than US GDP. Danish GDP is expected to have recovered in 2024. On the other hand, US GDP is not expected to recover (reach its pre-crisis growth trend) within the next ten years. This implies a larger negative effect of the pandemic on the US economy.
    • Fewer individuals have passed away in Denmark due to the pandemic. In late September, 0.06% of the US population had passed away. In Denmark, “only” 0.01% had passed away. This reduces the cost of premature mortality in Denmark.
    • The value of a statistical life is lower in Denmark. According to CS, in the US, it is USD 10m. In Denmark, it is USD 5m. This also reduces the cost of premature mortality in Denmark.
    • The numbers indicate that fewer people in Denmark will face health impairments, compared to the US. This also reduces the economic costs in Denmark, compared to the US.
  • CS find that the cost of the pandemic exceeds the cost of the financial crisis by a large margin. In 2012-2013, I chaired the official committee investigating the causes and consequences of the financial crisis in Denmark (link). We found that the financial crisis – at the time of writing the report in 2013 – had caused a DKK 200bn reduction in GDP. In 2008, when entering the financial crisis, Danish GDP was DKK 1,800bn, i.e., the accumulated loss up until 2013 amounted to 11% of pre-financial crisis GDP. Given that Danish GDP had not recovered to its pre-crisis trend in 2013, the final loss of the financial crisis is larger. It seems that the economic cost of the financial crisis and this pandemic will be, more or less, equal.

Uncertainties and assumptions

Many uncertainties surround the calculations presented above. In this section, I discuss some of them and their consequences. Notice as a starter, though, that by relying on the assumptions of CS above, I am able to compare results for Denmark with results for the US. If using other assumptions than CS, this is less straightforward. This – that it is easier to compare across countries when using the same underlying assumptions – is an argument for using the CS assumptions. Nevertheless, to see how sensitive results are, let us discuss what happens if using other assumptions.

Will Danish GDP have recovered in 2024, as the Danish Economic Councils expect? If not, the cost will be higher. If sooner, the cost will be lower.

Will 25 Danes pass away per week over the next year? This seems like a high number these days, given that we expect vaccines to be ready within the next few weeks. If “only” 650 (i.e. half the assumed number in the base-line calculations) additional Danes pass away, the cost of premature mortality will be reduced from DKK 71bn to DKK 49bn, and, all other numbers equal, the total cost will be reduced to DKK 300bn, or 14% of GDP. I.e., the baseline calculation is reasonably robust towards this adjustment.

The cost of mental health impairments in Denmark is associated with considerable uncertainty in this calculation, as we do not have official numbers for how Danes have been affected by depression and anxiousness during this pandemic. The overall baseline calculation is reasonably robust towards this number, though. For instance, I double the 2012 number in the baseline calculation. If I multiply with three, i.e. assume even more people are affected by mental health challenges, the total cost increases from 16% of GDP to 17% of GDP.

Finally, I, like CS, assume that the cost of premature deaths is 70% of the value of a statistical life. More than half – 52% – of Covid-19 related deaths in Denmark have been above 80 years old. Even when I, like CS, mark down the value of a statistical life by 30%, the number might still seem high. Reducing this further would of course lower the total cost of the pandemic marginally.

Conclusions and policy implications

This post presents first calculations of the expected economic cost of the Covid-19 pandemic in Denmark. Taking into account the direct effects on GDP, as well as the economic value of premature deaths and health consequences, I expect the economic cost of the pandemic to be 16% of Danish pre-crisis GDP. The calculation is based on a number of assumptions, but the overall figure seems robust to reasonably variations in these assumptions.

Harvard Professors Cutler & Summers expect that the cost of the pandemic in the US amounts to 90% of US GDP. The cost in Denmark will most likely be considerably smaller. Only a fifth of the cost in the US. This is due to a faster economic recovery in Denmark, a lower number of deaths in Denmark, a lower value of a statistical life in Denmark, and lower associated health costs. This large difference between the US and Denmark implies that there is considerable cross-country variation in the cost of the pandemic.

Cutler & Summers argue that the cost of this pandemic will exceed the cost of the financial crisis in the US. In Denmark, it seems that the cost of this pandemic will correspond, more or less, to the cost of the financial crisis.

Even when the cost of the pandemic in Denmark is considerably lower than the cost in the US, the cost is still enormous. 16% of GDP is a very large figure. For a family of four, it corresponds to DKK 240,000 (app. USD 40,000).

When the cost is high, it becomes even more important to roll out vaccines as fast as possible, as soon as health authorities have approved them. Call in retired doctors, call in retired nurses, use all available working personnel, and pay them all handsomely to work 24/7 in order to vaccinate as many as possible as quickly as possible, from a health perspective and an economic perspective. This will help contain the already very high cost of this pandemic.

VV or W: When did (or does) this recession end?

This corona recession started in February 2020. Officially, it is still ongoing. But, perhaps, it has in fact already ended. This might seem confusing but it helps explaining the performance of financial markets during this “recession”.

In my soon-to-be-released book From Main Street to Wall Street (link and link), I – among many other things – carefully examine the historical relation between the business cycle and financial markets. I verify that stock markets typically perform considerably better during expansions than recessions. In the book, I examine and explain why this is so. I also explain that this is not a bulletproof finding. It is not always so. Sometimes stock markets do fine during recessions. Is this recession one of them?

This recession

In the US, the Business Cycle Dating Committee (link) determines peaks and troughs in US economic activity. On June 8, 2020, the Committee announced that:

“The committee has determined that a peak in monthly economic activity occurred in the US economy in February 2020. The peak marks the end of the expansion that began in June 2009 and the beginning of a recession. The expansion lasted 128 months, the longest in the history of US business cycles dating back to 1854.”

This means that this recession started sometime in February 2020. It also means that the expansion preceding this recession became the longest on record.

The Committee has not declared the end of this recession yet. Officially, the US economy is still contracting, with the caveat that the Committee determines business-cycle turning points with a lag. E.g., it was in June only that the Committee concluded that the recession had started in February.

How has the stock market performed during this recession? As readers of this blog know, it has performed well since March 23 when the Fed had is “Whatever it takes moment” (link). During this recession, the US stock market has returned 10% up until today (total return on the MSCI USA index). This is particularly noteworthy in light of the fact that this recession has been unusually severe (link and link).

Perhaps the recession ended in April

This corona recession is special. Usually, recessions are caused by some economic imbalances that need to be corrected, such as an overvalued housing market or an implosion of the financial sector. This is not the case here. This recession was caused by the sudden arrival of a virus. The sudden arrival of the corona virus caused a sudden shutdown of economic activity.

The Business Cycle Dating Committee has traditionally relied on aggregate economic data when determining turnings points in the business cycle. Aggregate data, such as industrial production, GDP, etc., are available at the monthly or quarterly frequency. Given the underlying cause of this recession, which led to the closure of business activities from one day to the next, we need to look at higher-frequency data when searching for turnings points in economic activity.

One interesting new higher-frequency indicator is the Weekly Economic Index developed by the New York Fed (link). It summarizes information in ten weekly indicators of real economic activity. It is designed to track the four-quarter growth rate of real GDP. As the St. Louis Fed notices (link) ”This series is potentially a useful indicator to watch because James Stock (one of the authors of the study developing the index) is a member of the BCDC (Business Cycle Dating Committee).”

The interpretation of the index is that it is “scaled to the four-quarter GDP growth rate; for example, if the WEI reads -2 percent and the current level of the WEI persists for an entire quarter, one would expect, on average, GDP that quarter to be 2 percent lower than a year previously.” The Weekly Economic Index is available since January 2008:

FRED Graph
Weekly Economic Index for the US economy.
Source: Fed St. Louis Database.

The index tracks the financial crisis well. In particular, as the figure shows, it bottoms out in Q2 2009 when the recession ended.

The indicator also tracks the beginning of this recession well. It fell like a stone from a rock when the recession started in February.

The interesting point here is that the indicator reaches its bottom during the last week of April 2020. I.e., if a recession ends when this indicator bottoms out, as in 2009, this recession ended in late April.

The idea that this recession ended in late April would also be consistent with the observation that US real GDP dropped dramatically from the first to the second quarter of 2020, and rebounded strongly from the second to the third quarter:

Quarterly percentage changes in US real Gross Domestic Product.
Source: Fed St. Louis Database

What about international evidence? I am not aware of a well-designed index that tracks, e.g., Eurozone economic activity at the weekly frequency. So, as an alternative when looking at data from other countries than the US, let us consider industrial production. Industrial production is a traditional business-cycle indicator. It is available at the monthly frequency:

Indices of industrial production, normalized to one in January 2019, for the US and the Eurozone.
Source: Fed St. Louis Database and Thomson Reuter Datastream via Eikon.

For the US, industrial production bottomed out in April, like for the weekly indicator above. The collapse in Eurozone industrial production is much larger than in the US, but the timing is the same. Also for the Eurozone, industrial production bottoms out in April. So, perhaps this recession really ended in April.

The stock market and the recession

Typically, stock markets tank during the early phase of a recession, and starts rebounding when the end of the recession is in sight, i.e. before the actual end of the recession. Furthermore, the stock market normally does well during the early phases of expansions.

This graph splits the cumulative return on the US stock market into a mid-February (start of recession) to late April (assumed end of recession) period, in red, and a period thereafter, in green:

Total return index for MSCI USA, normalized to one mid-February 2020. Daily data. Red line indicates assumed recession. Green line indicates assumed early phase of expansion.
Source: Thomson Reuter Datastream via Eikon.

If we decide that this recession ended in late April, instead of assuming that the recession is still ongoing, the behavior of the stock market makes perfect sense. In this case, the stock market lost 14% during the recession. Since May, i.e. since the assumed end of the recession, the stock market has gained 27%.

Zooming in on the correlation between the Weekly Economic Indicator and the cumulative return on the US stock market, it follows that the stock market bottomed out before economic indicators started improving. This again emphasizes that the reason for the turnaround in the stock market in March was the “Whatever it takes moment of the Fed on March 23” (link). Since late April, however, the stock market and economic conditions have moved in tandem, i.e. economic conditions have improved and so has the stock market:

Total return index for MSCI USA and Weekly Economic Index. Weekly data
Source: Fed St. Louis Database and Thomson Reuter Datastream via Eikon.

Did the recession really end in April, then?

We do not know. It is, as mentioned, the NBER Business Cycle Dating Committee that determines peaks and troughs in the US economy. They have not called the end of the recession yet. Hence, officially, the US economy is still contracting.

I am happy that I am not a member of the NBER Business Cycle Dating Committee. This time, it must be particularly difficult to decide whether the US economy (as well as the European economy, by the way) is really out of the recession or not. As we all know, numbers of new cases are on the rise again in the US and have been doing so for a while in Europe. This will hurt economic activity going forward. So, should one conclude that there was a recession from February through April, or should one conclude that the recession is still ongoing. This is no easy question.

VV (two single Vs) or W?

Did the recession start in February and stop in April? Are we now entering a new recession due to a rising number of corona cases and a subsequent slowdown in economic activity? Or, are we still in the recession that started in February? One way to try to judge this is via NowCasts, i.e. daily forecasts of what growth in economic activity will be the current quarter. This is related to, but different from, the Weekly Index above. It is an estimate of the growth rate of GDP during the current quarter, but based upon data available at a higher frequency. Here is the NowCast from the New York Fed:

NowCasts of US quarterly growth in real GDP. Daily forecasts of growth during the current quarter.
Source: New York Fed

It has a clear V-shape during the second quarter. During April/May/June, incoming data indicated that the fall in GDP during Q2 would be enormous, as it turned out to be (see figure above with actual GDP growth). Incoming data during Q3 indicated that Q3 growth would be high, as it turned out to be. Q4 growth is expected to be considerably lower than Q3 growth, i.e. a new V, but growth is still expected to be positive.

Why is it important whether the recession lasted from February through April only (one V), whether we left the recession in April/May but enter a new now (two Vs, i.e. VV), or whether we are still in the recession, but economic growth was high during Q3 but expected to be low during Q4 and possibly Q1 2021, i.e. W? The reason is that it influences how we should think of recessions and financial markets during recessions.

Conclusion

Historically, stock markets have suffered during recessions. This recession started in February. Since then, the US stock market has gained 10%, seemingly in contrast to its usual behavior during recessions. Economic activity bottomed out in late April, though. The US stock market lost 14% from February through April. Since then, it has returned more than 25%. So, whether you conclude that the stock market has suffered during this recession or not depends on your favorite definition of when the recession ended, if at all until today.

To get the official answer, we have to wait for the NBER Business Cycle Dating Committee.

For the future path of the stock market, the question will be how the number of new infections develop and their impact on economic activity.

For the history books, i.e. for the conclusion of whether the stock market performed surprisingly well during this recession or not, the question is whether the recession ended in April or hasn’t ended yet. Did we have one V from February through April, do we get a new V now, so this ends up being VV, or are we still in the recession, i.e. W?

If Biden wins

If history is any guide, it will be four good years on the stock market if Biden wins on November 3. Historically, stocks have performed so much better under Democratic presidents. The question is whether history will be a guide also this time around.

Are Democratic or Republican presidents better for the stock market? To evaluate, let us recall the performance of the US stock market under Trump – a Republican – and compare it to its performance under Obama – a Democrat and Trump’s predecessor. Afterwards, let us look at the full history of the stock market under Democratic and Republican presidents. Finally, let us discuss what it implies for this election and the next four years on the US stock market.

Obama vs. Trump

This graphs shows the cumulative return to USD 1 invested in respective January 2009 (Obama 1st term), January 2013 (Obama 2nd term), and January 2017 (Trump). I show real returns, i.e. returns after inflation:

Cumulative real returns to US large-cap stocks under Obama and Trump. Own calculations.
Data source: http://www.econ.yale.edu/~shiller/data.htm

A couple of months remain, but during most of Trump’s presidency, the stock market has performed worse than under Obama. The exceptions are the first few months of Obama’s first term and the beginning of this year, right before the corona crisis.

In numbers, one US dollar invested in the US stock market in January 2009, when Barack Obama – a Democratic president – was inaugurated, had turned into 2.68 USD in December 2016 (when Obama’s second term ended) in real terms, given reinvestments of dividends. This is a total accumulated real return of 168 percent over eight years, or an average annual return of 13%.

One US dollar invested in the US stock market in January 2017, when Donald Trump – a Republican president – was inaugurated, had turned into 1.46 USD in September this year (in real terms). This is a total accumulated return of 46 percent over four years, or an annual average real return of 10% (there are still three months to go of Trump’s presidency, so the numbers might change a little in the end).

Some extraordinary events influenced the stock market under both Obama and Trump.

Early 2009, the stock market was still suffering from the financial crisis of 2008. Obama was inaugurated in January 2009. A few months later (in March 2009), the stock market turned around. Following March 2009, many great years on the stock market followed.

This year, 2020, has seen the fastest bear market in history (in March; link), caused by the corona crisis, hurting the stock market’s performance during Trump’s presidency.

The financial crisis of 2009 and the corona crisis were very unusual events. Perhaps the stock market’s performance under Trump and Obama was unusual.

It turns out that the picture painted above – that the stock market performs better under Democratic presidents – is robust. In fact, the US stock market has historically performed much better under Democrats.

The Presidential Puzzle

Pedro Santa-Clara and Rossen Valkanov published in 2003 a paper (link) entitled ”The Presidential Puzzle”. They documented an intriguing stylized fact: Over the 1927-1999 period, the excess return on the US stock market has been nine percentage points higher under Democratic than Republican presidencies. The average excess returns under Democratic presidencies, they showed, was 11 percent versus 2 percent only under Republican presidents. A nine-percentage point difference is enormous. For instance, it exceeds the excess return on the stock market in general, i.e. the return stocks provide over and above the return on risk-free assets.

Santa-Clara and Valkanov took great care in investigating potential reasons for this stylized fact. The most obvious explanation, financial economists would suggest, is that this is simply a compensation for risk, i.e. that risks have been higher under Democratic presidents. It turns out that this is not the case. In the end, their conclusion was:

”There is no difference in the riskiness of the stock market across presidencies that could justify a risk premium. The difference in returns through the political cycle is therefore a puzzle. ”

Pastor and Veronesi (link) update the results of Santa-Clara and Valkanov. They add app. twenty years of data (until 2015, i.e. their sample runs from 1927 through 2015). Pastor and Veronesi find even stronger results. In their extended sample, they find that the average return under Democratic presidents is eleven percentage points higher than the average return under Republican presidents.

Pastor and Veronesi have this figure with average annual excess returns under each president:

Average excess returns during individual presidencies.
Source: Pastor and Veronesi (2020).

The figure shows that the only Democratic presidency that delivered returns significantly below the average return over the whole period (the dotted line) was Roosevelt’s second term (1937-41). In contrast, a number of Republican presidents have delivered subpar returns (Nixon, first Reagan term, Bush Jr.). My small calculation in the beginning of this post, that returns have been higher under Obama than Trump, strengthens this conclusion.

Pastor and Veronesi show that this finding is robust during subsamples, for instance if excluding the large negative and positive returns during the 1930s. They also show that it is particularly the first year of the presidency that drives these differences. During the first year in office, Democrat presidents have experienced a 37%-points (!) larger return than Republican presidents have:

Difference between average excess returns in the returns during the presidents’ early years in office.
Source: Pastor and Veronesi (2020).

(If you are interested, here is a detailed account of the stock market’s performance under every single president since Truman: link).

There is some discussion whether this finding (that the stock market does better under leftwing presidents) holds internationally. Pastor and Veronesi show in the (114 pages…) appendix to their article that it holds internationally while others claim that it does not (link). We leave this aside here.

Democratic presidents are elected when times are bad

Pastor and Veronesi develop an interesting explanation of this intriguing pattern of the data. They argue that the reason for the difference between the stock market’s performance under Democratic and Republican presidents has nothing to do with economic policies during presidencies but everything to do with the economic situation when new presidents are elected.

Pastor and Veronesi argue that Democratic presidents are elected at points in time when risk aversion is high. Remember, for instance, the election of Obama in autumn 2008. This was right in the middle of the financial crisis. The worst financial and economic crisis since the 1930s. Everybody was afraid and uncertain how things would evolve. Risk aversion was high. When risk aversion is high, investors demand high compensation for taking on risks in the stock market. Subsequently, investors get compensated by high average returns.

So, the story of Pastor and Veronesi is that Democratic presidents get elected when times are bad. When times are bad, voters have a tendency to vote for left-wing candidates, as voters demand more social insurance during hard times. An important implication of this theory is that Democrat presidents do not cause higher stock returns (and Republican presidents do not cause lower stock returns). Instead, Democratic presidents are elected when the economy is suffering and risk aversion is high, causing high expected returns and subsequent high average returns.

Implications for this election

If (and I write if) it is correct that Democratic presidents get elected when risk aversion is high, and that good returns subsequently follow, what does this imply for this election and the next four years on the stock market?

Pastor and Veronesi suggest a number of proxies for risk aversion. To illustrate, let us look at two of them. The first is unemployment. The idea is straightforward: when unemployment is high, people are afraid of taking on risk on the stock market. Risk aversion is high and so are expected returns.

Here is unemployment in the US since 1947:

US unemployment rate.
Source: FRED.

Unemployment is currently high, at 7.9% in September (latest figure at the time of writing). This is more than two percentage points above the historical average rate of unemployment of 5.8%. Unemployment is coming down, though, and fast. Also, there are still two weeks to go until the election. Nothing is for sure. But, as it looks now, this indicates high risk aversion, and thus high expected returns.

Another measure of risk aversion is the habit-persistence idea of Campbell and Cochrane (link). If your consumption is low today, in relation to your past consumption, you feel that times are bad. If you are used to go on holiday a couple of times per year, but now you don’t dare because you are afraid of losing your job, you also become afraid of taking on risk in the stock market. Your risk aversion is high. You require high expected returns if you should be convinced to nevertheless invest in stocks.

Let me present a simple illustration here, inspired by Atanasov, Moeller, and Priestley: link. I take consumption (real per capita) and divide by habit. I approximate habit by the average of the past three years of consumption (this is a blog, so this should be OK), i.e. the habit ratio is here: C(t)/(AVG[C(t-1)+(C(t-2)+…..+C(t-12]), where C(t) is consumption in period t and AVG is the average. I do this calculation for every quarter since 1950. The result is the following time series:

Consumption habit ratio. Own calculations.
Data source: FRED.

When an entry in the figure is above one, consumption during that quarter is higher than its past three-year average, and times are good. When the habit-ratio is low, times are bad. The habit ratio (current consumption to habit) thus has a tendency to drop around recessions (1980, 2008, etc.), as, during recessions, people cut consumption in relation to past their consumption.

The lower is the habit ratio, the higher is risk aversion and thus expected returns. Currently, due to the enormous drop in consumption resulting from the corona-crisis, the habit-ratio is historically low. Times are bad. Consumption is low. Biden should win (according to this theory). Risk aversion is high. Expected returns should be high. The next four years on the stock market should be good.

There is uncertainty, of course

Many things can happen until November 3 and the next four years. And, for sure, this has been a really weird campaign. Nothing is for sure.

Some reservations:

My figure with the habit-ratio is based on quarterly data (consumption is quarterly). The last entry is Q2 2020. Since Q2, the stock market has been sprinting ahead, implying that the potential for future returns is, all else equal, lower now than in Q2.

Also, financial markets have been pricing in an enormous amount of event risk surrounding this election. The uncertainty surrounding this election is high. Protection against event risk is expensive (link, link, and link).

At the same time, the stock market has been doing really well since March 23 (link). This is unlike recessions in general. During normal recessions, the stock market tanks. This stock market rebound since spring, everything else equal, reduces the potential for future returns as of now.

On the other hand, at least right now, the polls seem to offer some support for the story. We are in a big downturn (corona crisis). In downturns, people have a tendency to vote left-wing. As we all know, Biden leads the polls at the time of writing:

Poll average. October 16, 2020.
Source: www.realclearpolitics.com.

Conclusion

The stock market has performed worse under Trump than under Obama. This is not an outlier. Historically, the US stock market has performed much better under Democratic presidents.

One explanation why the stock market performs well under Democratic presidents is that Democratic presidents get elected during bad times, when risk aversion is high.

Currently, we are in the midst of a terrible situation (corona crisis). Supporting this theory, Biden leads the polls. If history is any guide, this indicates good years ahead on the stock market. Not necessarily because Biden – if he wins – implements policies that support the stock market (perhaps he does, but this is not the point of the theory), but because risk aversion is high right now, at the time of the election.

Of course, so many things can happen. This certainly has been a strange (and, at least for many Europeans, myself included, a very weird and scary) campaign to follow from the sideline. The performance of the stock market during this recession has also been weird. It is important to recognize that the stock market rebound since March reduces the potential for future returns as of now. And, in general, so many things can happen during the next four years. Perhaps the situation will thus play out differently this time around. We will know how the election turns out in two weeks, and we will know how the stock market performs during the next presidency in four years and three months. It will be interesting to follow.

Expected returns, autumn 2020 updates

Today, October 1, the Council for Return Expectations publishes its updated expectations. We expect very low – negative – returns on safe assets over the next five and ten years. We expect an equity premium around 5-7 percent.

I chair the Council for Return Expectations (link). Twice a year, we update our expectations. Today, we publish our latest forecasts.

In a June post (link), I described the history of the council, how we operate, who we are, why we publish expected returns, what they are used for, and so on. Briefly, here, the Council consists of Torben M. Andersen (professor of economics, University of Aarhus; link), Peter Engberg Jensen (former CEO of Nykredit and current chairman of Financial Stabilitet; link ), and myself as Chairman. Based on inputs from Blackrock, J.P. Morgan, Mercer, and State Street (thanks!), we estimate expected returns, risks (standard deviations), and correlations between returns on ten different asset classes over the next five and ten years. We also publish expected returns on two assets classes (stocks and bonds) for investment horizons exceeding ten years.

The forecasts are important in Denmark. When Danish pension funds project how the retirement savings of their customers will most likely develop (with confidence bands), they base their projections on the Council’s return expectations. Similarly, when banks advice Danish investors on their non-retirement savings, they use the expected returns provided by the Council. I think it is fair to say that Denmark has been a front-runner in designing a system where banks and pension funds do not compete on expected returns, but all use a common set of return assumptions, determined by an independent Council.

The forecasts published today will be used for pension projections and projections for savings outside retirement accounts as of January 1, 2021. In spring 2021, we will publish updated expectations that will be used from July 1, 2021, and the following six months. And so on.

The timing is as follows. The expectations we publish today are based on market data from July 1, 2020. We use the period from July 1 until today (October 1) to determine our expectations. The expectations we publish today are then used by banks and pension companies as of January 1, 2021.

The numbers

We publish forecasts for expected returns on ten asset classes over the next five years, years 6-10, and over the next ten years. Returns are average annual returns in Euros/Danish kroner (the Danish kroner is fixed to the euro). These are our expectations:

Source: Council for Return Expectations

We expect an investment in safe assets (defined as a five-year duration portfolio of 30% Eurozone government bonds (minimum Triple-B), 20% Danish government bonds, and 50% Danish mortgage bonds – Danish mortgage bonds are triple-A rated, by the way) to lose money every year on average over the next ten years. This is noteworthy. On average, we expect such an investment to lose 0.1% per year over the next ten years. Over the next five years, we expect it to lose 1.2% per annum. The reason why we expect such low returns is that the underlying bond portfolio has a duration of five years, as mentioned, and we expect a normalization of interest rates, causing capital losses which drag down expected returns.

We expect lower returns from bond investments than we did in our June update (link). In June, we expected government and mortgage bonds to return a negative 0.3% per year over the next five years and a positive 0.3% per year over the next ten years. The reason why we expect lower returns now (-1.2% and -0.1% for the five and ten year periods, respectively) is that our forecasts today, as mentioned, are based on market data from July 1. On July 1, interest rates and credit spreads were lower than those used at our previous update. Our previous update was based on March 31 market data. In March, the corona virus had caused a hike in yields and a widening of credit spreads. Since then, yields and spreads have come down.

Global equities are expected to yield around 6% per year. This means that the equity premium is expected to be 5%-7%, depending on the investment horizon.

We expect low inflation. We expect Danish inflation to be 1.2% per annum over the next five years and 1.5% over the next ten years.

For investment horizons above ten years, we expect equities to return 6.5% per annum and bonds 3.5%. This is the same as we expected last year (we update long-run expectations once a year). Over the coming year, we will make a thorough evaluation of these long-run expectations.

We also publish standard deviations for all asset classes and for all horizons, as well as correlations and fees. Standard deviations are high for high-return assets; nothing comes for free. We thus publish many numbers. You can find all these numbers on the webpage of the Council: link.

Conclusion

We (The Council for Return Expectations) expect that Danish investors will lose money – even before fees and inflation – if investing in safe assets over the next ten years. You can expect positive returns on other asset classes, but then you need to take on risks.

Historically, the risk-free real interest rate has been around 2% per annum and the nominal interest rate around 4%. Investors would see their savings grow, even without taking on risk. This is not what investors should expect going forward. The fact that you cannot expect your portfolio to grow without taking on risk tells us a lot about the low-return world we are living in.