Worthwhile Canadian Initiative

Never mind the flatness, feel the length (of the observed Phillips Curve)

Sorry. It suddenly came to me this morning: a simple (and now blindingly obvious) way to reconcile an apparent contradiction in my own thoughts.

As a lot of people in different countries have noticed, the observed Phillips Curve now looks very flat. Certainly a lot flatter than it did in the past.

And I've been saying of course it looks flat; that's because central banks are now targeting inflation and were doing all sorts of different daft things in the past. The whole point of targeting 2% inflation is to try to make the observed Phillips Curve as flat as possible at 2%. And not just the Phillips Curve; whatever you put on the horizontal axis, if you've got inflation on the vertical axis the central bank is trying to make that observed curve flat. This point is starkest if we assume the central bank has rational expectations and is targeting its internal inflation forecast at 2% given a (say) 2 year targeting horizon. Deviations of inflation from 2% are then the central bank's inflation forecast errors, and if the central bank has rational expectations those forecast errors must have a 0% correlation with any variable or set of variables in the bank's information set 2 years prior.

If the central bank had an NGDP target instead of an inflation target, we might see the observed Phillips Curve having the "wrong" slope. If the bank foresees a period of higher than normal real growth and falling unemployment, it will try to make inflation come in below normal to keep NGDP growth on target.

It's a bit like the standard simultaneous estimation problem in econometrics: If you plot the correlation between P and Q, are you estimating a demand curve or a supply curve or a mix of both? The observed curve could slope either way. Except in this case it's like having a monopoly supplier of apples who adjusts the supply to whatever he thinks will keep the price of apples falling at 2% (the value of money falls at 2% if there's 2% inflation).

The flat observed Phillips Curve is just an artefact of the new monetary policy regime; a statistical illusion.

But at the same time I felt that something structural had indeed changed, and the Phillips Curve really had gotten flatter. The 2008/9 recession really did look like a standard recession caused by a negative shock to Aggregate Demand, and that recession lasted a longish time, and inflation should have fallen below target by more than it did, and certainly not stay the same on average, or even rise in some cases. Inflation targeting works well if "Divine Coincidence" is true; if inflation being above/below target coincides with output and employment being above/below where they should be from the point of view of monetary policy, so the central bank is too loose/tight. Deviations of inflation from target work as a good proxy for monetary disequilibrium, in other words. And Divine Coincidence seems to look a lot worse over the last 10 years than I thought it would look before the recession. Something structural did seem to have changed that made the Phillips Curve flatter, so the same amount of noise shifting the flatter Phillips Curve up and down meant inflation worked as a far worse proxy for monetary disequilibrium. Inflation had become much stickier. And maybe it was inflation targeting itself that had made inflation stickier, destroying its own signal of monetary disequilibrium. So we ought to switch to NGDP targeting instead.

But what evidence could justify my sense that something structural had changed, when a flat observed Phillips Curve is what we would expect?

It's not that the observed Phillips Curve is flat; that is unsurprising. It's that the observed Phillips Curve is flat and long; that is what is surprising.

It's the difference between flat and short, like this:

       xxxxxxx

And flat and long, like this:

x   x   x   x   x   x   x   x   x   x

where each x is an observation - a point in {inflation,unemployment} space.

Either the structural Phillips Curve really has gotten flatter, or else the variance of the natural rate of unemployment is a lot higher now than it used to be, and that does not seem plausible to me.

I feel much better now I've resolved my own cognitive dissonance.

Statistics Canada's historical housing cost data is wrong

In the early 1960s, Canadian economic historian Marvin McInnis started digging through the Dominion Bureau of Statistics archives, looking for city-level information on rental prices. While there, he discovered something strange and disturbing: 

A prominent theme of my career has been to reveal anomalies in what has been put forward as evidence. One instance is known only to me. In 1961 I was doing research for my doctoral thesis at what was then called the Dominion Bureau of Statistics. I wanted to know spatial price level differences across Canada. The data were collected for several major cities across the country but only changes over time for those cities were published as the CPI. I wanted geographical differences. I was on good terms with Simon Goldberg, the Assistant Dominion Statistician and he said , OK just go into the archive where the original reports from the cities were stored and dig out what you need. So I did.

What I discovered is that the clerks who recorded the data were under an instruction that, since the CPI was to represent prices paid by better off working class families, to edit out any rental figures what were above a designated threshold. By the end of the 1950s they were throwing out more than half of the reported rents. There had been a serious downward bias in a price that got fairly heavy weight in the CPI. I reported that to Simon but I don't think any alteration was ever made. It was water under the bridge. - Marvin McInnis, personal communication, November 23, 2016

Another Foray into Data: New Macro-Financial Data

I think Stephen Gordon's Project Link and its piecing together of fragments of Statistics Canada data is a solid step in the right direction.  If our national statistical agency is not going to provide long-term consistent data series, then I suppose its up to the researchers to lead the way.  Another case in point is the recent public release of a no-cost open access macro-financial data set by Òscar Jordà (Federal Reserve Bank of San Francisco, UC Davis), Moritz Schularick (University of Bonn, CEPR) and Alan M. Taylor (UC Davis, NBER, CEPR).

Do "Monetary Shocks" matter much?

The central bank is on the Gold Standard. The discovery of new gold increases the supply of gold, which causes an inflationary boom. People blame the inflationary boom on the gold discoveries. It can't have been a monetary shock, because the central bank wasn't doing anything different from what it always does. The central bank "did nothing". Gold mining needs to be a regulated industry to prevent this happening again.

Or a sudden fashion for gold jewelry increases the demand for gold, which causes a deflationary recession. People blame the deflationary recession on the fad for gold jewelry. It can't have been a monetary shock, because the central bank wasn't doing anything different from what it always does. Regulation is needed to control bubbles in jewelry markets.

In those two examples it's easy for us to see the fallacy. It was the gold discoveries, or fashion for gold jewelry, given the Gold Standard, that caused the monetary shock, which in turn caused the macroeconomic fluctuations. It's easy because we don't see the Gold Standard as "normal"; pegging the price of gold is not the way that we nowadays frame monetary policy. We don't use the price of gold as our measure of whether monetary policy is "tight" or "loose". And we are pretty sure that those two shocks would have had little macroeconomic effect if the central bank had a more sensible monetary policy than the Gold Standard.

But it might be a lot harder to see the same fallacy created by our own frameworks for thinking about monetary policy.

External validity, fallacies of composition, and the Wawa multiplier

The Canadian government decides to run an experiment to see if fiscal policy works. It throws 100 darts at a map of Canada. One dart lands on Wawa Ontario, so it spends an extra $1 million in Wawa. Local GDP in Wawa, and local GDP in all the other 99 places where the darts land, increases by $2 million relative to the control group of the Rest of Canada. The experiment clearly shows there's a government expenditure multiplier of 2.0.

Does that mean there's a government expenditure multiplier of 2.0 in the whole of Canada?

You should be able to see the underlying fallacy of composition. "Government spending has a multiplier of 2.0 in each part of Canada, therefore government spending has a multiplier of 2.0 in the whole of Canada."

Is it ethical to sell complimentary copies of textbooks?

Faculty Books Recycling is a company that takes the complimentary copies of textbooks that publishers send professors, resells those comp copies to students, and makes a profit on the transaction.

Faculty Books does everything possible to make professors feel that selling - or giving away - comp copies is an ethical thing to do. In their emails soliciting textbooks from faculty (sample below), they remind potential donors of the good that textbook recycling does. It puts textbooks into students' hands at a low price.  Those professors who choose to sell rather than give away their comp copies are told, "the money could sponsor a student event, be donated to  charity, or spent however you like."

Another rant about Statistics Canada's Attention Deficit Disorder

I've done this before, but I've had a very bad day and I need to vent.

I've started work on a forecasting project for the Canadian data, focusing on the short- and medium-term outlook. The first task is to see what sort of data are available, and once again, the frustrating answer is 'not near enough.' Statistics Canada's habit of starting new data definitions and not revising the historical series makes it extremely difficult to get a handle on what's actually happening to the Canadian economy.

Econometric estimates of fiscal policy multipliers

New Keynesian macroeconomics says that a (temporary) increase in government spending will cause an increase in the natural real interest rate. And unless the central bank increases the actual real interest rate by an equal amount, the result will be an increase in Aggregate Demand, which will cause an increase in real output and/or inflation. There is a fiscal policy multiplier. Plus, if the central bank holds the nominal interest rate constant, any increase in inflation and expected inflation will cause the real interest rate to fall, which will cause an additional increase in Aggregate Demand, multiplying the original fiscal policy multiplier. (In an open economy part of the effect works through the real exchange rate.)

This sounds sensible to me. It all depends on whether or not there is monetary policy offset. (Though, strictly speaking, it also depends on what exact goods the government buys; if for example the government buys goods that are perfect substitutes for private consumption or investment, which is like the government doing your shopping for you with your credit card, fiscal policy should have no effect.)

How to test whether fiscal policy actually works as advertised by New Keynesian theory? Macroeconometrics is hard.

The biggest problem here is controlling for monetary policy; because theory says that whether or not fiscal policy works depends on how monetary policy responds to fiscal policy.

Ontario and Its Neighbours

I suppose I have become somewhat obsessed with Ontario’s economy and its performance but here I go again with a few comparisons. Ontario is strategically located on the Great Lakes-St. Lawrence waterway adjacent to the huge population of the US northeast.  Its neighbours are trade partners and markets as well as economic competitors.  How does Ontario stack up to its immediate American and Canadian neighbors – namely, Minnesota, Wisconsin, Michigan, Ohio, Pennsylvania and New York, and then Manitoba and Quebec?

VAR vs WTF!?

Vector auto regressions (VARs) are supposed to tell us how the economy would respond over time if hit by a shock, by looking at past patterns of responses to shocks. A "shock" means "a deviation of one of the variables in the VAR from the level that was forecast by the VAR". And "shocks" include policy shocks.

1. Suppose at its next meeting the Bank of Canada increased the overnight rate from 0.75% to 1.50%. And suppose this was a totally unexpected move. And suppose a computer glitch meant that the Bank's statement explaining why it had done this was never published. How would markets react to the shock?

2. Now ask exactly the same question, but suppose the Bank of Canada instead reduced the overnight rate from 0.75% to 0.00%. How would markets react to the shock?

A VAR would give exactly opposite answers to those two questions. If you take the answer to the first question, reverse all the signs, you would get the answer to the second question. (That's only strictly true for a linear model.)

I think the answers to the two questions would be pretty much the same. The immediate response would be: "WTF!?".

The census long form - A critical tool for better understanding and policy

The following was written by Russell Davidson, Charles Beach and Thomas Lemieux on behalf of the Canadian Economics Association. A condensed version appeared in Saturday's Globe and Mail.

With a bill to reinstate the mandatory Census long form currently before Parliament, we wish to express our views, on behalf of the executive council of the Canadian Economics Association, on the importance of reinstating a mandatory long form for the 2016 Census.

Economists aren't in the prediction business - and that's a good thing

Last year one of my students* was trying to explain why immigrants struggle in the job market. His regressions weren't working, so he switched things round a bit.

Using 2006 Census data, he found that people in Newfoundland are 30 percentage points less likely to be immigrants than people in Ontario. People with PhDs are 21 percentage points more likely to be immigrants than people with a high school education. And people who are divorced are 9 percentage points less likely to be immigrants than people who are married.

He was crushed when I said "You can't use any of this in your paper. It doesn't make sense to have immigrant status as your dependent variable."

Reverse Regression and the Great Gatsby Curve

[Update: I'm getting an awful feeling this post may be totally wrong. Does (1/b)/[1+Var(V(i))/Var(Y(i))] = b if the Lamarck model is true? If so, my proposal won't work. Wish I could do math.] [Update2. Nope, I think it will work. But Majromax in comments has come up with a simpler way to do the same thing. See my 01:19 pm comment.]

Economists normally run a regression of son's income on father's income, and interpret the estimate beta as a measure of "intergenerational mobility". (High beta means low "mobility"). I think we could learn something interesting by reversing the regression -- run a regression of father's income on son's income -- and comparing the results of the two regressions.

This post is about: econometrics (errors in variables bias); genetics (regression toward the mean); and mobility/inequality. I don't know much about any of those three areas, so I will probably get things wrong. I am not 100% confident my proposal will work, but I would like people with better technical skills than me to take a look at it.

Ideally, we should all agree on how we would interpret the results before we run any regressions.

Will somebody please think of the Data Generating Process!

Or: "Should Finance People be allowed out unaccompanied by a macroeconomist?".

If the central bank is doing its job right, monetary policy ought to appear to be irrelevant.

The Economist says (HT Mark Thoma) "Interest rates do not seem to affect investment as economists assume", and this means that monetary policy is irrelevant. (Let's just set aside the fact that monetary policy is not interest rate policy.) The Economist refers us to an empirical study by S.P. Kothari, Jonathan Lewellen, and Jerold B. Warner.

How to test whether the LMI (LMCI) is a good indicator

The Bank of Canada calls theirs the "Labour Market Indicator". I now learn from Tim Duy (HT Mark Thoma) that the US Fed has one too, and calls theirs the "Labor Market Conditions Index". I think that LMI and LMCI are roughly the same sort of thing. It's an index number that is supposed to measure the amount of labour market slack [or tightness, for the LMCI], that is based on more than just the unemployment rate. (The Bank of Canada reports an LMI for the US too, and I don't know how much it agrees or disagrees with the Fed's LMCI.)

I read Tim as asking, quite reasonably, how we can know whether the LMCI is a useful indicator for the Fed to watch. Simon van Norden asked basically the same question about the LMI in a comment on my post. In particular, does the LMI/LMCI add any useful information for monetary policy that is not already in the unemployment rate?

Here's how to test whether the LMI is a useful indicator:

Dumb questions about econometrics and GAI/NIT

I've been meaning to write this post for some time. Scott Sumner's post spurs me to write it now.

First let me ask my dumb econometrics question. It's a very simple question, and I really ought to know the answer. But I don't.

Q1. If you estimate a linear regression, using Least Squares or whatever, will the estimated residuals always sum to zero? If the answer is "yes", then skip the next question. [Update: Ben, Norman, and Matthew tell me in comments the answer is "yes", provided there is a constant term in the regression, and I want there to be a constant term.]

A simple test of simple rules against actual policy in the actual economy

A "simple rule" is a formula that tells a central bank how to set the nominal interest rate as a fixed function of a small number of variables. The Taylor Rule, which sets the nominal rate as a function of the gap between actual and target inflation, and the gap between actual and potential output, is the classic example of a simple rule. Are simple rules better than discretion? Here's how we can do a test to find out, and the results of an old test where simple rules failed to do better than actual policy.

Fiscal Policy Shocks

Well, here is a new contribution to the debate over the effect of fiscal policy shocks from the journal of the Institute for Fiscal Studies.  The authors Paweł Borys, Piotr Ciżkowicz and Andrzej Rzońca are from the Warsaw School of Economics and look at the impact of fiscal policy shocks on EU new member states.  Their results will be seen as contribution to the debate over austerity measures and whether one should stimulate economic activity through tax reductions or direct expenditure measures.

Flogging a Dead Horse III: The weird response of Benjamin Zycher

This post was written by Mike Veall of the Department of Economics at McMaster University.

Sorry, I am still on about the research that was used to formulate the Ontario Progressive Conservative jobs plan. Again I emphasize this is about the underlying econometric research by Benjamin Zycher, Senior Fellow of the American Enterprise Institute. It is not about what I think the plan’s proposals actually will or won’t do.

Why do beginner econometricians get worked up about the wrong things?

People make elementary errors when they run a regression for the first time. They inadvertently drop large numbers of observations by including a variable, such as spouse's hours of work, which is missing for over half their sample. They include every single observation in their data set, even when it makes no sense to do so. For example, individuals who are below the legal driving age might be included in a regression that is trying to predict who talks on the cell phone while driving. People create specification bias by failing to control for variables which are almost certainly going to matter in their analysis, like the presence of children or marital status. 

But it is rare that I will have someone come to my office hours and ask "have I chosen my sample appropriately?" Instead, year after year, students are obsessed about learning how to use probit or logit models, as if their computer would explode, or the god of econometrics would smite them down, if they were to try to explain a 0-1 dependent variable by running an ordinary least squares regression.

I try to explain "look, it doesn't matter. It doesn't make much difference to your results. It's hard to come up with an intuitive interpretation of what logit and probit coefficients mean, and it's a hassle to calculate the marginal effects. You can run logit or probit if you want, but run a linear probability model as well, so I can tell whether or not anything weird is going on with the regression."

But they just don't believe me.