How QE works

Various commentators are saying that Europe’s QE won’t work, or QE in general doesn’t work because it just boosts the value of assets. Increasing the reserves of banks, critics say, doesn’t cause banks to lend money to the real economy.

That’s irrelevant. QE is not supposed to make banks lend more money. Banks don’t need reserves to lend money, or rather it works the other way. Banks lend money if there’s demand for loans, and then ask for reserves which are always given.

What QE does is indeed to boost asset prices. Central banks buy bonds, people who sold the bonds buy stocks, stocks go up in value. Or people who sold the bonds spend money, money ends up in company profits, stocks go up in value.

And this how QE works. What happens when stocks go up in value? Companies expand and hire more people. What happens when stocks fall in value? They cut costs and lay off people. When stocks rise in value pension funds are wealthy. When they fall, poor.

In our imperfect system QE is a blunt instrument that makes rich people richer while boosting the economy. The problem, though, is with concentration of financial wealth, not with QE.

A short critique of the Efficient Market Hypothesis

The so-called Nobel prize in economics has been awarded to Gene Fama for his Efficient Market Hypothesis. EMH states that markets instantly price in all available information and so nobody would be able to outsmart or otherwise outperform the market consistently in the long run.

Here’s my short critique of the theory.

The EMH assumes there’s an event horizon. Events behind the horizon, be they unknown future events or insider secrets, have no bearing on prices. Then events pop up over the horizon and instantly the market computes new asset prices to fully reflect the new information. Ergo you can’t out-compute the market.

The assumption that the market computes prices instantly is idealistic but we’ll go with that. The market is pretty fast, down to seconds or less.

Real-world events don’t flip from fully unknown unbiased probability to fully known outcomes. There’s a bias i.e. any particular event is predicted as more likely to happen than not, or vice versa, and better confidence estimates of the event’s probability become available over time. But so long as we assume everyone has access to the same stream of predictions the EMH still makes sense.

Where the EMH falls down is that prices don’t change to reflect the final valuation of a future state as soon as that future is known. New information gets priced in over time, from when the information is revealed to the time when the new situation actually takes effect and directly bears on the fundamentals. People see the instant tick of the pricing and say “ha, EMH!” but there’s a lot more pricing yet to come, and that’s why prices change continuously even in the absence of important news.

The reason markets price in information over time is twofold:

  • Market participants have different trading time frames. If we know for certain that the US will default in one year that will cause stocks to drop instantly, but there’s still time to invest and get out during the year, so people do. If the time frame is uncertain there’s more scope for price change. Miscalculations about trader’s ability to enter and exit cause bubbles to inflate and then crash.
  • Since the market isn’t pressured to price in the impact of future events until the exit window of each type of trader closes, it doesn’t, and how it will eventually price the impact over time remains unknown. Market participants have to predict prices at specific times and and the analyst with the better prediction of the market’s reaction wins. For example if you and I predict that a default will cause a 5% or 30% drop in asset prices tomorrow (not eventually when the default happens) one of us will come out looking smarter.

So, even in a world where everyone has full access to information about events, including likelihood and confidence, there are still opportunities. Opportunities arise from being better or worse at estimating how the market will compute price changes over time given known inputs, which is a notoriously hard but valid computational problem.