Friday, January 5, 2018

Moving Blog Location

New Year, new address and look for my blog. The new address is http://antoniofatas.blogspot.com

If you have been subscribing to this blog, you can change your subscription by going to this new link.

Antonio Fatás

Thursday, December 29, 2016

Make the Risk Premium small again

In one my previous posts I looked at the stock market valuations in the US to conclude that they were in line with recent historical data. In fact the stock market looked cheap relative to most years since the mid 1980s.

But that was before the US election! Since the election the stock market has gone up and interest rates have gone up as well. How do stock market valuations look like today?

I follow the same methodology of the previous post and start with the Price-Earnings ratio constructed by Robert Shiller. We know that this P/E ratio has been high relative to historical averages (today it stands above 28, a level only achieved before in the 2000 bubble or in 1929.

But the P/E ratio depends on several macroeconomic variables, in particular the level of real interest rates. In my previous post I made the argument that once we correct for the level of the real interest rate the current P/E ratio looks reasonable.

In particular, if we express the price of stocks as the net present (real) discounted value of earnings. Under the assumption that current earnings are expected to grow (in real terms) at a rate G and using R to denote the risk-adjusted discount rate we can write:

P = E / (R-G)

In other words the Price-to-Earnings ratio can be written as

P/E = 1 / (R-G)

To make the expression easier to read let's invert it

E/P = R-G

And let's express the risk-adjusted discount rate as the sum of a risk-free rate (RF) and a risk premium (RP).
E/P = RF + RP - G

This expression says that the E/P ratio is a function of three factors. Each of them can make the ratio low (i.e. stock prices high relative to earnings): either real interest rates are low, or investors expect earnings to grow fast or they feel good about risk and they are willing to accept a low risk premium.

The last two terms are the ones that depend on expectations, can be more volatile and are capturing the optimism or pessimism of investors regarding macroeconomic conditions (both potential growth and perceptions of risk). While there is nothing in this formula that allows us to assess how irrational investors are, one would expect that at times where stock prices are seen as "bubbly" are when either estimates of growth are too high or the risk premium is too low. Let's combine the two together by rewriting the equation above to construct a potential "bubble" index of the stock market

"Bubble" Index = G - RP = RF - E/P

Measuring the real risk free rate using 10-year interest rates and inflation expectations from forecasts of inflation from the survey of professional forecasters posted at the Philadelphia Fed we can get the updated picture below.
















We can see that after the election the index has increased both because of higher stock prices and also higher interest rate. The stock market is more "bubbly" than back in October but still in line with historical estimates.

One issue that might be relevant for this chart is that we have been downgrading expectations of potential growth over all these years. How much would our assessment change if we took this into account?

Using CBO potential growth estimates for the following 10 years as a measure of estimated real earnings growth we can calculate the implicit risk premium as:

RP= E/P -RF + G

This number tells us the implicit compensation that markets expect for risk given current E/P ratios, real interest rates and assuming markets share the views of the CBO about future growth rates (and that earnings grow at similar rates as GDP).

Below are the results of such calculation. 
















What we can see in this chart is that today the risk premium remains in line with historical estimates although it has come down after the election, as expected. The risk premium reached a very low level at the end of the 1990s when we combined extremely high P/E ratios (above 40) with fairly high real interest rates. This is strong sign of a bubble because either markets overestimated growth (relative to CBO forecasts) or underpriced risk.

The risk premium was extremely high during the 2008-09 crisis when P/E ratios were extremely low and interest rates had come down very fast, but those were times of very high uncertainty. In the last coupe of years the risk premium has been around 6%. While stock prices climbed, interest rate fell and kept the risk premium stable.

Since the election, the risk premium has gone down by 1% and it is about 5%. That is not a low number but the change is significant. Clearly markets are either:

a) upgrading their expectations of growth relative to CBO estimates
b) or assuming the policies of the new administration will make earnings grow faster relative to GDP
c) or feeling that uncertainty has gone down with the election of Trump as the next president

How much growth will increase remains a source of debate. While some believe that potential expansionary fiscal policy might boost growth, it is still unclear how much of it will be implemented and the actual policies that will be adopted might not be as growth friendly as markets believe. Given what we know so far I remain skeptical about the potential for acceleration of growth rates in the coming years.

Earnings could grow faster than GDP growth with policies that are friendly to companies but as the labor market becomes tighter it is not obvious how large this effect can be.

And this takes us to the last issue: risk. Clearly investors are not pricing much risk in the stock market, quite the contrary. This is a surprise. Regardless of what you believe about Trump's stated policies, he has provided very few details on what his actual policies will look like, he has contradicted himself in numerous times and some of his statements open the door for very damaging events from a political and economic point of view that could lead to a global economic slowdown or a recession (not to mention other, more catastrophic, events). The market is clearly not pricing any of this risk. I find this puzzling.

Is this a bubble? We will find out soon but it starts looking like the beginning of one where two very dangerous features are present:

a) investors associate high P/E ratios with potentially higher stock market returns because they extrapolate recent trends. This is wrong. High valuations mean lower future returns (if everything else remains constant)
b) investors seem to be underpricing risk at time where some cautions might be needed

We are still 1% point from the risk premium at the 2007 peak and very far from the 2000 craziness so there might still be some room to go. But let's remember that those are the years that preceded a significant fall in the stock market, not the right benchmark to use.

Antonio Fatás


Tuesday, November 29, 2016

The OECD procyclical revision of fiscal policy multipliers

The OECD just published its November 2016 Global Economic Outlook.  Their projections suggest an acceleration of global growth rates in particular in countries with plans for a fiscal expansion. In the case of the US, and based on the "plans" of the Trump administration, the OECD projects an acceleration of GDP growth to 3% in 2018.

I am very glad to see that the OECD is more open to the idea that a fiscal expansion might be the right policy choice in a low growth environment. I am also very happy that they are ready to admit that fiscal policy multipliers are larger than what they previously thought.

But I am puzzled that they seem to ignore their previous disastrous economic policy advice. And I am even more puzzled that they are upgrading their estimates of fiscal policy multipliers (in particular for tax cuts) at the wrong time in the business cycle, when the economy must be closer to full employment.

Here is the history: back in 2011 many advanced economies switched to contractionary fiscal policy at a time where their growth rates were low and unemployment rates remained very high.  During those years the OECD seemed be ok with fiscal consolidation given the high government debt levels (consolidation was necessary). They understood that there were some negative effects on demand but as they assumed multipliers or about 0.5 (in the middle of a crisis with very high unemployment rates!) the cost did not seem that high.

Today, in an economy with unemployment rate below 5%, and wages and inflation slowly returning to normal vales and a central bank ready to raise interest rate, the OECD turns around and decides to change the fiscal policy multipliers to something close to 1 even if the announced fiscal measures consists mostly of tax cuts to the wealthier households with low propensity to consume.

This is what I would call a procyclical revision of fiscal policy multipliers. Encourage consolidations in the middle of a crisis and expansion in good times. Not quite what optimal fiscal policy should look like.

And, of course, the media (including the Financial Times) reported on the OECD study as a validation of the new US administration policies.

And I leave for another (longer) post the absence of any serious discussion of the risks associated to a Trump presidency. This is coming from an organization that has been obsessed with the risks of inflation and excessive asset appreciation during the crisis.

Antonio Fatás

Monday, October 17, 2016

The stock market looks cheap

How expensive is the US stock market? Have the strong gains since the financial crisis of 2008 built yet another massive bubble that will require a correction? Some investors fear that recent stock market record levels are a sign that this is the biggest bubble ever. Of course, such a simplistic comparison of stock prices is flawed both because stock prices are nominal and therefore likely to go up over time and also because they are driven by (real) earnings which are also likely to increase as (real) GDP increases.

To correct for these trends we can do a simple adjustment and look instead at the price-to-earnings ratio. I will use here the one constructed by Robert Shiller (although some ague that recently the index could be less informative because of the way earnings are being calculated).
















The P/E ratio shows that the end of the 90s bubble was by far the period were stocks looked the most expensive relative to earnings. What do P/E ratios look like today? On the expensive side. With a ratio above 26 it stands right at the level before the 2008 crisis and a lot higher than previous similar historical episodes. Most tend to compare it to 16, as the average P/E ratio in recent decades, to signal that the stock market is very expensive. Without going back many decades, we could say that the stock market today looks as expensive as it has been since 1981 with the exception of the bubble of the late 90s.

But that cannot be the end of the analysis as we know that the P/E ratio depends on several macroeconomic variables, in particular the level of real interest rates. And we know that real interest rates are at very low levels today and likely to say low for a long period of time. How does the above analysis change if we adjust for changes in the interest rate?

Let's go back to the basic finance equation that links the P/E ratio to macroeconomic fundamentals. Start with a simple expression of the price of stocks as the net present (real) discounted value of earnings. Under the assumption that current earnings are expected to grow (in real terms) at a rate G and using R to denote the risk-adjusted discount rate we can write:

P = E / (R-G)

In other words the Price-to-Earnings ratio can be written as

P/E = 1 / (R-G)

To make the expression easier to read let's invert it

E/P = R-G

And let's express the risk-adjusted discount rate as the sum of a risk-free rate (RF) and a risk premium (RP).
E/P = RF + RP - G

This expression says that the E/P ratio is a function of three factors. Each of them can make the ratio low (i.e. stock prices high relative to earnings): either real interest rates are low, or investors expect earnings to grow fast or they feel good about risk and they are willing to accept a low risk premium. The last two terms are the ones that depend on expectations, can be more volatile and are capturing the optimism or pessimism of investors regarding macroeconomic conditions (both potential growth and perceptions of risk). While there is nothing in this formula that allows us to assess how irrational investors are, one would expect that at times where stock prices are seen as "bubbly" are when either estimates of growth are too high or the risk premium is too low. Let's combine the two together by rewriting the equation above to construct a potential "bubble" index of the stock market

"Bubble" Index = G - RP = RF - E/P

What this equation says is that the stock market is going to look expensive when the risk free rate looks high relative to the earnings-to-price ratio, given that the only way to justify such high stock prices is through expectations of high growth or perception of unusually low risk.

What does this index look like? I constructed the difference between RF - E/P by using price-to-earnings ratio and 10 year nominal interest rate from Shiller. And I converted nominal into real interest rates using forecasts of inflation from the survey of professional forecasters posted at the Philadelphia Fed  (I mostly use 10-year inflation forecasts that match the duration of the nominal interest rate. But for the earlier years only the 1-year inflation forecast is available to I have decided to plot both series, which looks almost identical).

This is what the stock market "bubble" index looks like
















This chart tells a very different story from the unadjusted P/E ratio. The 90s bubble is still there, as you would expect because interest rates were high. This means that the very high stock prices during those years could only be justified by very optimistic views on growth or perceptions of unusually low risk. Both were present and they turned out to be wrong, that's what a bubble looks like.

We can also see that the financial crisis of 2008 sent stock prices close to the lowest levels, similar to what happened during the double recession of the early 80s. Since then stock prices have recovered but they remain low relative to any of the previous years. Even if we ignore the 90s as an aberration, compared to the expansion of the 80s or the 2000s, the stock market today remains "cheap". And by cheap we mean that to justify current prices we do not need a very optimistic view on growth or that investors are demanding a very high risk premium. In other words, the stock market tells us that either investors are pessimistic about growth or very risk averse (which is the opposite of what you expect to see during a typical bubble).

Does this mean that the stock market is undervalued? No, it all depends on whether our current growth expectations or risk assessment are correct. Growth might surprise us and be even lower than what we think today, risk could be a lot higher, maybe 10 year interest rates are a really low indicator of what interest rates are going to look like in the next 10 years. In all these scenarios the stock market will look more expensive than what it looks today. But growth could also surprise us through gains in productivity. And it might be that some of the current risks do not look that likely a few years from now. And then the stock market will look really cheap. That's the uncertainty that we always have when trying to assess the pricing of the stock market.

A note on real growth expectations: our view on growth has clearly been revised downwards because of unfavorable changes in demographics and productivity. It is possible that future real GDP growth today is 1% below what it was in the 80s or the 2000s. But even if we were to make that correction to our index, the stock market would not look expensive. If you look carefully at today's index the number is about 1.5% lower than in the 80s and about 2-3% lower than in the 2000s. These differences are larger than revisions to potential GDP growth rates.

In summary, unlike the strong warning signals we get when looking at record-level nominal stock prices or even at the P/E ratio, a simple adjustment of P/E ratios by current levels of interest rates paints a very different picture of the stock market. Adjusted for current levels of real interest rates, P/E ratios tell us that the stock market today is on the cheap side relative to previous similar phases of the business cycle.

Antonio Fatás


Tuesday, August 9, 2016

You can lower interest rates but can you raise inflation?

Last week the Bank of England lowered their interest rates. This combined with previous moves by the ECB and the Bank of Japan and the reduced probability that the US Federal Reserve will increase rates soon is a reminder that any normalization of interest rates towards positive territory among advanced economies will have to wait a few more months, or years (or decades?).

The message from the Bank of England, which is not far from recent messages by the Bank of Japan or the ECB is that they could cut interest rates again if needed (or be more aggressive with QE purchases).

Long-term interest rates across the world decreased even further. The current levels of long-term interest rates have made the yield curve extremely flat.













And in several countries (e.g. Switzerland) interest rates at all horizons are falling into negative territory.



















The fact that long term interest rates is typically seen as the outcome of large purchases of assets by central banks around the world. In fact, many see it as a success of monetary policy actions.

But if monetary policy is being successful we expect inflation expectations and growth expectations to increase. Both of these forces should push long-term interest rates higher not lower!  Something is fundamentally not working when it comes to monetary policy and it is either the outcome of some forces that the central banks are unable to counteract or the fact that central banks are not getting their actions and communications right.

On the communications I will repeat the argument I made earlier: When central banks repeat over and over again that they can lower interest rates even more they are misleading some to believe that lower interest rates (long-term and short-term, real and nominal) are a measure of success of monetary policy. This is not right. Lower nominal interest rates across all maturities cannot be an objective when inflation and growth are seen as too low. Success must mean higher nominal interest rates. And success must mean at some point a steeper yield curve not a flatter one.

Why are central banks failing in their communications? I see two reasons:

1. They want to send a message that they are both powerful and not out of ammunition. Repeat with me: "Interest rates are low and they can get even lower."

2. These are interesting times. With short terms rates stuck around zero, all the action of the yield curve has to come from long-term rates and, in addition, QE and the massive purchase of assets is also a new phenomenon that is not always well understood by market participants.

My guess is that it is this combination of circumstances that are unusual by historical standards and the difficulty of communicating a complex monetary policy strategy by central banks that are sending long-term interest rates to even lower levels. These levels are not consistent with any reasonable scenario for growth or interest rates over the next decades. When 30 or even 50 year interest rates are negative or close to zero something is not right. Either this is the end of growth as we know it or the start of a 30-year period of extremely low inflation combined with deflation or our expectations are seriously off and we are up for an interesting surprise.

Antonio Fatás

Wednesday, August 3, 2016

Experts, facts and media

Jean Pisani-Ferry has written a very interesting post about the need for trusted experts in a democracy. The post addresses the criticisms that economic experts have received as a result of the Brexit vote. Quoting from his post:
"Representative democracy is based not only on universal suffrage, but also on reason. Ideally, deliberations and votes result in rational decisions that use the current state of knowledge to deliver policies that advance citizens’ wellbeing."
Very well said. He also brings up the point that the lack of influence of economic experts is not that different from that of other experts (as illustrated by the debates on climate science, GMOs,...). I share that view and my guess is that the mistrust of economic experts is simply more visible because of their influence (or lack of) in the political debates that tend to be a lot more present in the media than the debates on scientific issues.

How to enhance the trust on experts? Not obvious, according to Pisani-Ferry. What is needed is a combination of of discipline among the community of experts, an education system that equips citizens with the tools to distinguish between facts and fiction and the development of better venues for dialogue and informed debate.

Good luck! Unfortunately we are very far from this ideal scenario. Education has reached more citizens than ever, more so in advanced economies, but we see little difference. It might be that the complexity of the issues that are being debated is at a level which still does not allow us to have an informed discussion based on facts and not ideology. Opinions that are expressed using either the wrong facts or no facts at all somehow are able to reach the public and have an influence that is as large as that of those who present the facts. And the media does not serve at all as a filter, maybe because controversy sells or maybe because there is a need to present a 'balanced' view of the debate or simply because of self-interest.

Here is my example of the day that illustrates this point: the Financial Times published two articles yesterday on the merits of quantitative easing. One argued for more QE under the logic that is working and we just need to increase the dosage. The second one presented the view that quantitative easing (as well as expansionary fiscal policy) are the wrong tools to use to generate a recovery and that they are likely to lead to a very unhappy ending.

If you read the second article you will notice the use of dubious facts and an economic logic that anyone who has ever taken any economics course should realize that is badly flawed.

Let me pick one example. The article starts with the figure of 300% of GDP for global debt and then it argues that
"If the average interest rate is 2 per cent, then a 300 per cent debt-to-GDP ratio means that the economy needs to grow at a nominal rate of 6 per cent to cover interest."
This is just wrong on so many counts:

  • The increase in debt in the world is matched by an increase in assets. 
  • The interest rate paid by borrowers goes to lenders. So the world (or a given country) does not need to find income to pay for this interest, this is a transfer from borrowers to lenders.
  • Borrowers need to pay for the interest but if debt is coming from a mortgage to buy a house, I do not need to pay rent anymore. Looking at interest payments alone (or at liabilities without taking into account assets) is just wrong.
  • The 300% number cannot be associated to a country or a government, most debt is internal. No country has an external debt that is anywhere close to that level. Same is true for governments (with the exception of Japan which is not far, but, once again, most of this debt is internal - so the interest that the government of Japan has to pay goes to the Japanese citizens who happen to be the taxpayers).
  • Even if you had a government that had 300% of debt, the calculation above is simply wrong. If interest rates are 2%, you need to grow at 2% (not 6%) to ensure that the debt-to-GDP ratio stays constant (as long as your additional borrowing or saving is zero, of course). This is something that is taught in a principle of economics course. The authors are confusing the value of interest payments and the required growth to make that level of debt sustainable.
The rest of the article contains many other mistakes. It is embarrassing that the Financial Times is willing to publish such a low quality article.

Will this article influence anyone's view on the debate on monetary policy? I do not know but what I know is that the pessimistic view presented in the article on the role that monetary and fiscal policy is popular enough that is still influencing both the debate around and also the outcome of current economic policies.

So we are very far from having informed and factual debates about the economic (and scientific) issues that shape economic and social outcomes. As an economist I continue to do my best by sharing my views and analysis with a wide audience through blog posts like this one but it is depressing to see how those that rely on flawed analysis often manage to reach the public through the validation of the most respected media.

Antonio Fatás


Wednesday, June 8, 2016

9, 8, 7, 6.7,... Speculating on China growth

The deceleration of China's GDP growth rate has been seen both as a natural transition towards more sustainable growth rates and a sign that the Chinese model of growth is coming to an end. How does this deceleration of growth rates compared to similar historical episodes for other countries? Is 6-7% a sustainable growth rate for China?

Let's frame these questions in the traditional model economists use to look at growth rates of emerging and low-income economies: the convergence model (based on the work of Robert Solow). The main prediction: countries that are lagging have more opportunities for investment and they are likely to grow faster than countries at the technology frontier. Because of faster growth rates we expect to see convergence in GDP per capita. As convergence happens, growth rates will naturally slowdown to reach those of the countries at the frontier. [The theory also states that not all countries might converge to the same level of GDP per capita but let's ignore that for a second and focus on the predictions of relative growth rates.]

Start with an example that nicely validates the theory: South Korea. Let's plot the level of GDP per capita of South Korea relative to the US at the beginning of each of the last three decades and then compare it to the growth rate of GDP per capita during the years that followed [I will treat the 2000-2014 period as the "decade of the 2000s"].



The plot above (click on it for a larger picture) shows that South Korea starts in 1980 with a level of GDP per capita below 20% of the US leading to annual growth during the 1980-90 period of about 7%). This allows the country to reach a level of 30% relative to the US by 1990. At that point growth is decelerating and during the next 10 years it goes below 5.5%. The transition continues with decelerating growth during the next decade (around 4%). Today South Korea has a GDP per capita of about 65% of the US level and it is likely that over the coming years we will see a further slowing down of its growth rate as it continues its convergence towards the US.

So far so good for our theory. How about China?



In 1980 China starts at a much lower level of GDP per capita and, consistently with our logic, a very fast growth rate of about 7%. But as the country converges towards the US growth rates accelerate to above 8% during the 90s. One potential explanation for this acceleration is that China was moving towards a more natural growth rate given how low its GDP per capita was relative to, say, South Korea. But the next decade  (2000-2014) will bring yet one more increase in growth rates towards 9%. At that point the growth rate looks spectacular compared to the case of South Korea. To put it in perspective, by 2015 China has reached a level of 25% of US GDP per capita (the vertical red line) and when South Korea had reached that level it was already growing at less than 6%. The comparison to South Korea makes clear that growth rates of 8-9% in China given its current development would have looked like a true miracle.

What if South Korea is used to benchmark future Chinese growth rates? Given the current GDP per capita of China, we are looking at a growth rate of slightly below 6% over the next decade. And a growth rate that will decelerate further as time passes. This number is slightly lower than the current target of the Chinese government, but not far given that we are talking about 10 years in a path that is likely to be one of decelerating growth rates. [Note that the growth rates above are in per capita terms. Working age population is currently not growing much in China so the GDP figures should not be too different].

Final question: is South Korea a good benchmark for China? Let's look at other potential fast-growing economies during the same years.




South Korea is clearly the best performer in the range of countries that are below 50% of the US GDP per capita (and, yes, there are plenty of failures!). So using South Korea as a benchmark is providing a very optimistic case on Chinese growth. There are a few other countries that look like outliers (from above) in this relationship but it is unclear that they are relevant examples for China. Hong Kong and Singapore are small city states. Ireland in the 90s is a very unique decade for a European country, and oil producing countries (such as Norway) have dynamics that cannot be replicated without that level of natural resources.

In summary, the deceleration of GDP growth rate in China seems like a natural evolution of the economy as it follows its convergence path. Growth rates around 6% still put China as the best performer among all countries in the world, conditional on its level of GDP per capita. There are plenty of other countries in the world that show China what low growth rates of GDP really look like. And they might be an example of what could happen to the country if it cannot keep its policies and institutions among the best in class for its current level of development.

Antonio Fatás