If you’re an ambitious political thinker, one of the things you should do is imagine a utopia. Almost everyone does it. Milton Friedman did it with his book Capitalism and Freedom where he listed what his ideal society looked like and steps to get there. Karl Marx did it, but was short on empirics and his Manifesto didn’t exactly have a well defined path to get to his ideal society. Stepahnie Kelton’s The Deficit Myth is another example of a book that promises to get a utopia.
One yet untried utopia is futarchy proposed by the economist Robin Hanson. Futarchy is a form of government where voters decide on objectives (maximise GDP, minimize inflation, increase life expectancy) but prediction markets decide the means to get there.
For example, if a country was deciding whether to cut taxes it would issue two securities: one which paid the holder GDP if taxes were cut, and one which paid the holder GDP if taxes remained the same. Then you could compare the prices of the two securities and see in which case GDP would be higher and provide information. Or as Hanson’s manifesto puts it
"Futarchy" is an as yet untried form of government intended to address such problems. In futarchy, democracy would continue to say what we want, but betting markets would now say how to get it. That is, elected representatives would formally define and manage an after-the-fact measurement of national welfare, while market speculators would say which policies they expect to raise national welfare. The basic rule of government would be:
When a betting market clearly estimates that a proposed policy would increase expected national welfare, that proposal becomes law.
Like many utopias, this is ambitious. It aims to replace current institutions of elected legislatures with betting markets and speculators decide what’s good or not. If you are part of the group of people who think that the only good financial innovation was the ATM, you won’t be happy with this idea. But keep an open mind, prediction markets have an exceptionally good record.
The Iowa Electronic Market which predicts electoral events is consistently more accurate than polls. A prediction run by the pharmaceutical company Eli Lilly to predict whether its drugs would be approved by the FDA or not was more accurate than its internal experts. Hewlett Packard’s internal prediction market on computer sales performed better than its own forecasts. A prediction market on the reproducibility of psychological studies outperformed the estimates of psychology experts. The US Department of Defense used prediction markets for predicting future political events, and the CIA had a detailed overview of their uses. Along with that, orange juice futures in the US are far more accurate in predicting Florida weather (where most oranges are grown) than the US National Weather Service.
There are two main reasons why prediction markets have done better than experts. The first is incentives and the second is a diversity of views.
Incentives are clear here. When experts make predictions, they do not face large rewards for getting it right or penalties for getting them wrong. On the other hand, traders can make large amounts of money if their predictions are right or lose large amounts if they are wrong. This gives them an added incentive for them to get their predictions as accurate as possible.
A diversity of views is another important benefit of prediction markets. It is usually hard to become an expert. It takes years of education and research. While this means that experts are intelligent, it also means that the risk of them being stuck in groupthink and not considering alternative perspectives exists. Experts can and have been stuck in their own groups without considering alternative views. Sometimes this leads to disastrous consequences when we rely on their judgement for decisions that affect millions of people. An example of where experts got things wrong and prediction markets got it right is the Federal Reserve in 2008.
Inflation futures
The USA has a market for predicting inflation. You can see it here where it predicts inflation for the next 5 years and here where it predicts inflation for the next 10 years. It's not strictly a market for predicting inflation. Rather, the US Government issues bonds whose principal changes with the Consumer Price Index. The coupon payments on these bonds are linked to the principal (they’re typically a percentage of the principal), and so when inflation goes up the coupon payments go up and vice versa. They’re called Treasury Inflation-Protected Securities or TIPS
The yield on these bonds is a measure of how much investors are willing to pay for putting their money in a safe asset with no risk of inflation or default. This is the real bond yield. Along with this you have normal Treasury bonds which pay a fixed amount every year. The investors on these bonds aren’t compensated for inflation, they have to bear the risk of inflation. This is the nominal yield. According to the fisher equation, the nominal rate = real rate + inflation. So, if you do subtract the real yield on TIPS from the nominal yield on an equivalent Treasury bond you get the expected inflation rate.
On 15 September 2008, the investment banking giant Lehman Brothers filed for bankruptcy. It is today, 13 years later, the largest bankruptcy in the world. Lehman’s collapse sent shockwaves in the market. Financial markets froze, and the S&P 500 dropped 4.96% on the day. Given the collapse in asset prices and increase in unemployment you might imagine that the Fed was forecasting a drop in inflation. But the exact opposite happened. The Fed forecasted an increase in inflation to 3% despite most of the inflation coming from a temporary price increase in commodity prices.
But at the same time, the bond market expected differently. In the same week of Lehman’s bankruptcy the market’s expectation of inflation dropped from 1.5% to 1%. This isn’t a small change - it showed that the bond market expected inflation to be much lower and that the Fed should have cut rates faster. But the Fed ignored the signal from the bond markets. Only 3 weeks later when the bond market was predicting only 0.62% inflation, the Fed decided to cut interest rates by 0.5% - which given the low rates of inflation and surging unemployment was too little. And even then despite the bond and stock markets signaling extremely low inflation due to reductions in aggregate demand, on October 29, Ben Bernanke the Fed’s Chairman argued that the Fed did its best.
In the end, the bond market was right and the Fed was wrong. The Fed’s preferred measure of inflation, the PCE Price Index, fell from 2.2% in August 2008 to 0.9% in July 2009 - which was the lowest it had been since the US government began tracking the index in 1960. The Consumer Price Index fell, and the US faced deflation for the first time since 1950. This was in line with the bond market’s prediction and the Fed was wrong at the time it mattered the most.
In 2008 it was a win for the markets and a clear loss for the experts
Given the clear success of markets during the worst financial crisis in the last 90 years you might think that institutions have started to accept prediction markets more. But you’d be wrong.
GDP linked bonds - solve the information problem
It only seems natural to extend the market approach to inflation prediction to another important variable - GDP. It is obvious that long term fluctuations in GDP are important to wellbeing. Countries that experience rapid economic growth end up better off after than before on a variety of indicators like life expectancy, personal incomes and infant mortality.
But what matters to us here is short term GDP changes - will your GDP next year be higher and lower than this year. And by how much? Having a market based estimate of this information would be extremely helpful. To begin with it would provide a real time, day by day forecast of GDP. And given that prediction markets have had a good run so far, it will probably be more accurate than conventional forecasters. So the question comes: how do you structure a security to predict GDP?
One easy response is to have a GDP futures market. The US Treasury could agree to create a security that paid the holder a dollar for every trillion of US GDP. So if GDP next quarter would be $22.567 trillion, the holder would get paid $22.567. (Or you could multiply it by multiples of 10 to get a number with fewer decimal points).
This is a nice first step, but it has some obvious problems. The first is the time value of money and the other is illiquidity.
If there was a GDP future for 30 years later, that would amount to the Treasury agreeing to pay the holder an unspecified amount of money in 30 years. This is very similar to a 30 year bond, except that it has no coupon payments in the middle, and the final value is determined 30 years later. You could do a lot of good things with that money! It could have gone in a bank account, been invested in stocks, in venture capital and so on. So demand for this will probably be low in the absence of interest payments in the middle.
The second problem of illiquidity remains. If you can’t trade this and sell it when you need it, there isn’t going to be a large market for buying these securities. And liquidity is a self fulfilling prophecy. If investors believe nobody will buy it from them when they need to sell it, they won’t buy the security. This leads to more illiquidity and so the whole process starts again.
A good way to avoid both of these problems is to have government debt payments indexed to GDP. This is a GDP linked bond. If interest payments are linked to GDP, then interest payments change as the economic fortunes of a country change. Along with that, you get a market for predicting GDP. This gives you the futarchy aspect of GDP linked bonds.
Where can you use them?
If we did have a GDP linked bond - what could happen? The first thing is that government interest payments would be countercyclical. When the economy did well, they would pay more in interest and if the economy did poorly, they’d pay less in interest. This would mean that in downturns, they’d save money on interest payments and it would allow for more fiscal spending.
But that’s not what this post is about. The advantages of GDP-linked bonds are well known, and there are books, research papers and blogs about it. See the References for more. But what’s ignored in these is the informational aspect of GDP linked bonds.
The first obvious benefit is that they will be extremely useful in a crisis for understanding the damage done. Imagine if, in March 2020 there was a market for estimating GDP losses in real time. This would have been a welcome addition to the cacophony of forecasts about GDP growth because it would be with people with skin in the game.
The second is that it would be able to change the incentive structures of politicians. You might complain that politicians focus too much on short term benefits at the expense of long term growth. Or that they get too much money from lobbyists and don’t care about ordinary citizens. Paying politicians in non-transferable long term GDP linked bonds can alleviate that. If a politician who was in control of economic policymaking was paid in GDP linked bonds which they could not sell (or sell with a predetermined schedule - like vesting schedules of tech company employees), then they’d have their personal economic futures aligned with the economic futures of ordinary people.
The third is that it would be able to create prediction markets for policies with GDP bonds as the base. For example, a country was deciding to implement a new tax policy. It could create a futures contract that predicted the price of the GDP bond conditional on the tax policy passing, and a futures contract for the price of the GDP-linked bond if the tax policy did not pass. So, then futures on GDP-linked bonds could be a way of implementing futarchy.
Who should issue them?
In theory, the countries that should issue GDP linked securities are developing countries. They typically have volatile GDP growth and pay higher rates of interest. They should be the ones issuing GDP linked bonds. But in practice that’s not so. To begin with, normal developing country bonds aren’t very liquid. If a developing country issued an untried, untested bond like this, they probably would not have many people willing to buy it. Because of the bond being illiquid, the implied GDP growth rate wouldn’t be very accurate.
But if a developed country with a liquid bond market could issue a GDP linked security, then it would set a precedent for other countries also to do it. An obvious example is the US where almost half a trillion dollars of bonds are traded daily. If the US were to issue a GDP linked bond, it would probably get a great deal of interest from investors. Hundreds of analysts around the world track US GDP growth and the market for other bond which tracks the US economy (TIPS) is over $1.5 trillion. So it's not unrealistic to expect that GDP linked bonds will have a great deal of interest from institutional investors.
Why hasn’t this happened yet (or why will this fail?)
(To be completely fair, GDP linked securities have been issued in Costa Rica, Bulgaria, and Bosnia and Herzegovina in the 1980s and 1990s; and since then, Argentina, Greece, and Ukraine. But they were warrants where holders got a bonus if GDP hit a certain level)
A good question to this is: if GDP linked bonds are so beneficial, why hasn’t this been done yet? Or another question is: how can this go wrong? A few hypotheses below
There is low demand for a security that tracks a country’s GDP
Government treasuries (and their debt management offices) are conservative, and haven’t innovated
GDP figures will be manipulated to reduce interest payments
I think number 3 is wrong. To begin with, governments have a large incentive to over-report GDP figures to boost their reputations, and the benefit of reporting lower GDP numbers would be far lower than the reputational cost of having lower GDP. Along with that in the inflation linked bond market where governments actually do face incentives to underreport data to reduce both reputational damage and public perception of inflation there is no evidence that they have been doing so.
Number 1 is the most damaging. But as I explained above, it's unlikely that this will happen (at least for the developed world) because investors already have shown interest in TIPS. To support my case here’s an op-ed from a group of large institutional investors calling on governments to issue them.
Number 2 is the most convincing option left. The last innovation in government bonds was the inflation linked bond which was started in 1981 by the UK, but took till 2007 till the rest of the developed world adopted it (that is the UK did it first in 1981, then Australia in 1985, Canada in 1991, Sweden in 1994, the US in 1997, France in 1998, Japan in 2004, Germany in 2006 and Japan in 2007) As you can see, it takes a lot of time for debt management offices to change their minds . Along with that, a government debt management office has almost no incentive to innovate, and politicians aren’t concerned with this either.
References
Books
Sovereign GDP-Linked Bonds: Rationale and Design https://cepr.org/content/new-ebook-sovereign-gdp-linked-bonds-rationale-and-design
Papers
GDP-linked Bonds: Some Simulations on EU Countries: https://ec.europa.eu/info/sites/default/files/economy-finance/dp073_en.pdf
Blogs/News
GDP-linked bonds: why so few, and why so expensive? https://blogs.lse.ac.uk/businessreview/2021/04/23/gdp-linked-bonds-why-so-few-and-why-so-expensive/
Letter: The moment has arrived for the GDP-linked bond - FT
https://www.ft.com/content/1d91548c-71db-11ea-ad98-044200cb277f
Financing the pandemic rescue package - The Hindu
https://www.thehindu.com/opinion/op-ed/financing-the-pandemic-rescue-package/article31292615.ece
It’s Time for GDP-Linked Bonds - Barrons
https://www.barrons.com/articles/its-time-for-gdp-linked-bonds-1521817236
Better still, the Fed could issue such a security and then target it's price to rise at 1+ its preferred inflation rate times 1+it's estimate of maximum real income growth. BTW the Treasury ought to be issuing more TIPS at other intervals 6 mo, 1,2,3,4 year intervals
These arguments seem to apply just as much to bonds linked to real GDP. So why is it better to make a futarcy optimising nominal GDP than one optimising real GDP?