Models, Data and US Federal Reserve Policy Decisions

 

One January 11, 2024, the PBS News Hour interviewed Raphael Bostic, president of the Atlanta Federal Reserve Bank. One interesting part of the interview was the description of how models vs. data are used in the Federal Reserve decision making process, at least by Dr. Bostic. The role of economic models in decision making is interesting because on July 20, 2010 the US House of Representatives Committee on Science and Technology held a hearing on the topic (transcript here) and concluded the models were not very useful. My conclusion is that the US Fed is on the right path with some of the new models being developed. Future posts will give my recommendations (as a statistician) and my reasons for making them. It is essentially my forecast for the future of US Fed model building.

In the interview above, Amna Nawaz asked when Bostic expected the economy to reach the US Fed's 2% Inflation rate target. Bostic answered:

Well, we have models, and models will give us an answer...[but]...I don't put too much stock in any of those longer-term issues...I just try to keep an eye on where things are going month to month and try to just have a clear understanding about where we stand.

In other words, the models give us some long term predictions about Inflation and Economic Growth but, for month-to-month decision making we use data.

So what exactly can we get from models? I would hope that that the Atlanta Fed gets Prediction Intervals telling them that, say, Inflation might be between 1% and 3%, bracketing he 2% target. Then, as the data come in, it can be evaluated to answer Anna Nawaz's question. And, I would hope that the state of the economy would have some role in predicting the time path of inflation. Something like the prediction intervals produced by Climate Models:

The original topic of the News Hour interview was to gage the actual strength of the economy and consumers expectations about economic growth (although most of the interview concentrated on Inflation).

 For the strength of the economy, let's look at the Atlanta Fed's GDPNow forecast for economic growth (quarterly percentage change in real GDP). The output from the GDPNow app (presented above) compares the GDPNow forecast to the range of the top and bottom ten Blue Chip forecasts. GDPNow predicts GDP percentage changes well outside the Blue Chip forecasts until we get into March of 2023. What's going on here and why is this happening?

Maybe it would help to look at a longer time period. The St. Louis Fed publishes the GDPNow output from 2014-2024 (above and here). We can very clearly see the COVID shock, the economy's response, and the return to approximately at 2% growth rate. Note that it took approximately three years to recover from the COVID shock.

There is a lot to scratch your head about in the NewsHour interview and the outputs of the GDPNow model before we even get to thinking about the problem of inflation. Why don't the Blue Chip forecasts show the COVID shock? Why does the GDPNow cellphone App not go back to 2019, before COVID, to report results? And, which forecast should we believe, if any? 

It sounds as if, from Dr. Bostic's comments, that the FED ignores the forecasts and just waits for data to come in when making decisions about the economy. That's OK, but the FED spends a lot of time and money on large-scale, Dynamic Stochastic General Equilibrium  (DSGE) models (here), the models criticized in the Congressional Hearings, models that produce yet another set of forecasts. Worse yet, the DSGE models are based on assumptions that economic agents use models to form expectations about economic variables and use these expectations to make decisions, decisions that DSGE models attempt to predict.

But, which models specifically are economic agents using: the GDPNow model, the consensus of the twenty-or-so Blue Chip forecasting models, the forecasts of the DSGE models, or some other model entirely (I have my own models that are similar to but, I argue, an improvement over the GDPNow approach). I know the Fed is trying to be transparent and lay everything out on the table but what I'm looking at appears contradictory as it must have looked to Congressional Committees. And,  some commentators (here) and Congressmen (here) want to get rid of the Federal Reserve, Fed forecasts and Fed policy manipulations entirely.

Interestingly enough, the current problems with Economic Policy all point back to our failure to understand the Great Depression* and the effects of economic shocks (such as the WWI-WWII shocks and the COVID shock). In future posts, I'll try to untangle this mess** because I think it is interesting and important but not because I think any economic agents (to include the Fed and the ECB***) will be interested. Eventually, I will get around to looking at Inflation and Deflation!

Notes

* ChatGPT (here) lists the following causes for the Great Depression: (1) Stock Market Crash of 1929, (2) Bank Failures, (3) Reductions in Consumer Demand, (4) High Tariffs and Trade Barriers, (5) Monetary Policy Mistakes, (6) Debt Deflation, (7) Decline in International Economic Activity and (8) The Dust Bowl and Agricultural (Environmental) Collapse. 

** My working hypothesis is that we need to embed the US Economy within the World-System to not only understand the Great Depression but also to understand current economic policy confusions. The Fed doesn't really have a role for the World-System in its models.

*** The failure of Macro-economic models was also felt by the European Central Bank (ECB): "Macro models failed to predict the crisis and seemed incapable of explaining what was happening to the economy in a convincing manner".

Why the US Government Can't See Economic Bubbles

In yesterday's NY Times, Jared Bernstein (former chief economist to Vice President Joe Biden) wrote a perspective-from-experts piece (here) titled "Undoing the Structural Damage to Potential Growth". Dr. Bernstein presumably gets the attention of policy makers in Washington and in this piece he aligns himself with the Congressional Budget Office (CBO)  for support. What is striking about the piece is that, without saying as much, he completely rejects the ideas that there was a Subprime-Mortgage Bubble which became visible in 2008 and argues that government policy could get us back to the peak level of GDP that was generated by the bubble. Let me try to summarize his argument.

The CBO (in the graphic above) has made two forecasts for "potential output," that is output the economy could maintain under full employment. The first forecast (bold dashed line) was from January 2007, right before the Subprime Mortgage bubble burst. It showed accelerating GDP growth into the foreseeable future. In February 2014, the CBO modified their forecast to show a lower rate of potential GDP  growth, 7.3% below the old forecast line which Bernstein translates into 10 million jobs lost. The Federal Reserve has argued that this labor market damage cannot be undone. Bernstein argues that it could if we ran a "high-pressure" economy fueled by government spending.

The theory underlying all this is an analogy derived from the physical sciences. When a physical system undergoes a permanent change (say the parts start wearing out because the system has been in production for a long time) the effect is referred to as hysteresis. Bernstein describes how the analogy works for economies:

When a cyclical problem morphs into a structural one, economists invoke the concept of hysteresis.  When this phenomenon takes hold...the economy undergoes a downshift that lasts through the downturn and well into the expansion, reducing the economy’s speed limit. 

The bottom line of his argument is that hysteresis can be reversed by "high-pressure" economic policy. He claims that results by "a number of top economists" are about to be published confirming the relationship between fiscal policy, hysteresis and reverse-hysterists.

To me, none of this argument is very clear: (1) If hysteresis is a permanent change how can it be reversed by spending more money rather than replacing the system's parts, the parts that are wearing out? (2) What is it about the Subprime Mortgage crisis that you would call cyclical? This is the worst recession since the Great Crash of 1929. That's a pretty long cycle. (3) What precisely is the mechanism that changes a cyclical problem to a structural problem (but it wasn't a cyclical problem)? What exactly is a structural problem, the parts wearing out? (4) Is the economy "downshifting" because the transmission is wearing out and gears are slipping or because the economy exceeded some speed limit (bubble)? OK, I'm really confused, but maybe this makes sense to Washington policy makers.

What's really going on here underneath all the analogies and mixed metaphors? Let's go back to the original CBO "potential output" forecasts. They use some basically simple equations to make these projections. If employment is a function of output then L = e(GDP), that is, employment is proportional (e) to GDP. If we reverse the equation and we assume some number for full employment, (L*)/e = GDP*, where L* is the assumed full-employment labor force and GDP* is full-employment output. This exercise is really just equivalent to picking some historical date and then drawing lines on graph paper--anyone can do it, not just the CBO.

In the graphic above I've plotted real GDP (GDPC96) over time (the solid black line). The dashed lines are the dynamic attractor paths for US GDP. I've written in more detail about how this plot and the attractor paths above were generated (here and here). The attractor path was basically generated by a state-space model of the US economy. It clearly shows the Subprime Mortgage Bubble and shows that we are currently pretty much on the attractor path in 2014. For understanding what the CBO is doing, you do not really need a model or even a simple equation.

You can pick any historical date and start drawing some lines. If you think 2003 was a reasonable place for the US economy to be, just connect it to another low point such as 2000 and you have a forecast out to 2009. If you think 2008 was a reasonable place for the US economy to be, as does the CBO and Jared Bernstein, then draw line B going off into infinity. If you changed your mind after the Subprime Mortgage Crisis, as the CBO did, draw line C and hope it returns to line B at some point in the future. Or, if you think 2012 was a reasonable GDP level for the US economy, draw line D which corresponds pretty well with the attractor path generated by the USL20 model.

Does anyone else beside the CBO and Jared Bernstein think that 2008 was a reasonable level for GDP? If you do, you do not believe in economic bubbles. You cannot take action to pop bubbles because bubbles do not exist. If you accept the Federal Reserve's argument that the Labor Market damage from the Subprime Mortgage Crisis cannot be undone and your common sense tells you that there was a Subprime Mortgage Bubble, the time to act was sometime between 2004 and 2006. The problem is that among those who accept the common sense idea of economic bubbles, no one can agree on how to identify a bubble (Line A or the USL20 attractor path) so policy makers cannot act. And, those who do not believe in bubbles are left with lines B, C and a comforting theory of potential output--or the harsh reality of line D.

Is the "Settled Science" Argument a Straw Man?

The recent publication of the Fifth Scientific Assessment (AR5) from the IPCC concluding that human being are altering the climate has been met by a uniform response from the Right Wing: the idea that Science is settled, whether it be Climate Change or Evolution is a Myth. The argument surrounding "Settled Science" has actually been going on for a long time, especially between the NY Times and the Wall Street Journal (here). If the Right Wing doesn't like a scientific conclusion or theory it just argues that the science isn't settled. The Left Wing argues that science is never settled.

A recent opinion piece by Charles Krauthammer in the Washington Post (here) uses the settled-science myth to criticize the current Administration in Washington DC for worrying about climate change or requiring that health insurance cover mammograms. Since a recent large-scale, randomized clinical trial showed little benefit from mammograms (here), Mr. Krauthammer suggests that Climate Science will shortly also be overturned and, what is more, most scientific conclusions can be ignored by political commentators.

It is unfortunate that both sides have chosen to use the idea of "settled science" because the idea is wrong. Science is not about results and conclusions. Science is about "models" and the evidence that accumulates for and against the models. Some models have better support than others.

The best supported scientific model of climate change can be summarized with the I=PAT Model above (sometime called the Kaya Identity and a reasonable simplification of complex Integrated Assessment Models). Population growth (N) leads to more economic production (Q) which leads to greater energy use (E) which leads to greater CO2 emissions which leads to increases in global temperature (T). The competitor model (if the Right Wing can be said to have such a thing) would be that global temperature is a random walk,  T(t) = T(t-1) + U. Tomorrow's global temperature is today's global temperature plus random error, U (unknown).

The random walk competitor model is easily defeated (here). Until someone comes up with a better model and until there is some evidence either for or against that unknown model, the I=PAT model is the best one we have. Arguments about "settled science" do not lead to better models.

The same arguments can be applied to Mammograms, PSA screening for prostate cancer, or any area of scientific interest. First, we have to ask if good models are available. In the case of many medical findings, good models are not available. There are currently few good models for the causes of cancer. Screening is an attempt to find something early before it progresses. Screening would be better if it was less intrusive and if we knew what we were looking for (BRCA1 genetic screening is one example). None of this has anything to do with settled science.

Why the "Free" Market Needs Regulation

The Triangle Shirtwaist Factory fire (March 25, 1911) was one of the deadliest industrial disasters in US history. Lest we forget why industrial capitalism needs government regulation and why there are labor unions, you can view the entire PBS American Experience documentary here.

Causal Model Recap for 2012

This blog is partly about developing causal models to understand the world around us as it is happening in real time (the other part is simply about me venting steam while I read the popular press). It is based on reading Judea Pearl's book Causality: Models, Reasoning and Inference.  Prof. Pearl is a computer scientists studying artificial intelligence by asking how we can formalize for machine learning what humans do so easily: establish causal connections between events. Prof. Pearl's answer to the question is through the use of directed graphs.

One way to test Prof. Pearl's ideas is to look at current events and try to clarify arguments by developing causal models. Looking back on a year of doing this kind of testing here are the models, brief descriptions of the arguments and a link to the original blog postings.


Arguably, the most important event in 2012 was the unfolding Financial Crisis that started in 2007. A model I developed in January of 2012 (here) looked at the role of inequality (inequality has been increasing the US during the period of Neoliberalism to levels not seen since the Great Depression in the 1930--see the graph here) and the effect that a Wealth Tax might have on economic growth. The model indicates that inequality has a role in decreasing economic growth and that a wealth tax would have the opposite effect. The model is meant to confront the right-wing argument that inequality is necessary for economic growth.


I developed another model on a similar topic in February (here). In this model (click to enlarge) I pointed out that right-wing seems to make its arguments by reversing the actual direction of causation. The argument is that the Entitlement Society is creating economic problems while the direction of causation is in the other direction: economic forces (financial crises and globalization) are creating the need for increased Federal benefit spending. More data (and more models) will be needed to determine whether reverse-causation is right-wing strategy or simply confusion over the direction of arrows in positive feedback loops (viz., warm temperatures causing people to emit more CO2).

In May I picked up the topic of climate change (here) looking at a recent argument being made by the Climate Denial crowd, in this case Richard Lindzen. He was arguing that global temperature increase resulting from CO2 emissions (notice that the role of CO2 emissions in climate change has been admitted here) would trigger a negative feedback effect that would keep global temperature under control.

The negative feedback control loop he hypothesized involved the reduction of Cirrus Clouds. Since Cirrus Clouds are thought to play a role in creating the greenhouse effect (increasing global temperature) any reduction would act to control temperature.

This is an interesting and sophisticated argument from the Climate Change Deniers. There are many poorly understood feedback loops in the global climate system (some are reviewed here) and Cirrus Cloud formation is certainly one of them. In this case, there is little data that supports the argument.

The world climate system is obviously complex and some parts (local weather) are probably chaotic (more here). For us non-climate scientists, our best hope is to be able to develop the arguments as causal models and watch as the data accumulates.

In November (here) after Hurricane Sandy I looked at the role of climate change in severe weather formation and its consequences. The causal model above shows that the difference between air and sea temperature caused by climate change, in addition to increase atmospheric water vapor, will increase the intensity and consequences of hurricane flooding.

In a later post the same month (here) I added economic causes to the model showing how economic growth creates not only increased CO2 emissions but also increased coastal development. With greater coastal development and greater CO2 emissions, damage from Hurricanes can be expected to increase even more.

I would like to promise that I could develop causal models for the entire climate system at some point this year.  That would be really useful but also pretty premature. It will be along time before enough data is available to use in critiquing the models. On the other hand, the IPCC is scheduled to finalize the Fifth Assessment Report (AR5) in 2014. It would be useful to have a collection of models ready to use in reading AR5. And of course, the Subprime Mortgage Crisis is winding down and will continue to provide opportunities for casual modeling as various political parties and commentators try to put the monkey on someone else's back for the event.