Thursday, December 20, 2012

Observations on economic forecasting

I decided to build a macroeconomic forecasting model. This is a strictly statistical model--it doesn't impose a lot of economic theory on the data. For readers who know a bit of statistics, it's a Bayesian VAR model with a Normal/inverse Wishart prior, basically patterned after this one.* (Many thanks to Saeed Zaman for a lot of helpful guidance on this--he explained some of his own work in this area, including a working paper on which my model is based, and he pointed me to a bunch of good BVAR resources. This is very similar to the exercise I described here, which was also done by Zaman.)

I'm going to talk about the model a bit then offer some observations about the problem of economic forecasting generally. Each section of this blog post is independent, so skip the ones that look boring to you (yes, I know that will probably mean you skip them all). I think that the section of the most general interest is the last one.

The model

For those unfamiliar with this kind of modeling, the setup is basically this: choose a handful of important economic variables (as dictated by forecasting needs and theory), collect a bunch of data on them (in my case, quarterly data from 1964 to the present), then use a statistical method to figure out the (linear) relationships that exist (a) between different variables and (b) within each variable over time. That basically describes VAR (vector autoregression) models.

It turns out that standard VAR models are dismally poor forecasters because they are too complicated (too many parameters); they can provide a really good fit of existing data, but they catch too much random movement in variables and aren't able to predict averages out of the data sample (they have an "overfitting" problem). Modelers have been able to improve them by using Bayesian methods. Specifically, Bayesian methods allow the modeler to, essentially, bias the model towards parsimony and simplicity (using the word "bias" loosely). Then the data have to be really informative to move the model away from that simplicity.

In my model, I start with the prior view that each variable is a "random walk" (with drift)--that is, each variable just moves randomly, maybe with some trend, and none of them influence each other. I also impose priors that set the bar progressively higher on how much past values of variables can influence present values. (for those who know, the prior mean is a random walk for each variable, and the prior variance for each coefficient shrinks as the lag increases--this is a shrinkage estimator, which is common in the BVAR literature).

To put it simply, my model considers movement and relationships among variables to be random unless convincingly shown otherwise. This kind of parsimony has been found to significantly improve forecasts.

Some model results

How good is my model? Imagine Woody Harrelson's voice asking, "compared to what?"

This may sound sad to some non-economists, but a common benchmark for evaluating economic forecast models is whether they can forecast better than a random walk--a model that literally specifies all variables as moving randomly. The standard way of doing this is to estimate your model on just part of your data, then use it to forecast the other parts of your data, and see how well the forecasts do by getting the average error (squared). Then compare that error to the error of a random walk with drift. I did this and looked at results for four variables: GDP, unemployment rate, inflation (core PCE), and the Fed's policy interest rate. I made some tables, but the one person still reading this will leave if I post them, so I'll just say this: my model beats a random walk model for GDP forecasts up to 6 quarters ahead, unemployment forecasts at least 8 quarters ahead, and inflation at least 8 quarters ahead. It fails miserably at forecasting the Fed's policy rate; this is pretty standard in models like this (for understandable reasons I won't go into).

By the way, the random walk with drift approach isn't as silly as it sounds. If I had been living in a cave for a few years and someone demanded that I forecast U.S. real GDP for next year, I'd probably say "2.5 to 3 percent higher than it was this year." (The variables are logged, so the drift term is the average growth rate).

I can also compare my model to a standard VAR--that is, a very similar model that does not use any kind of Bayesian approach. My model handily beats the standard VAR for all variables at all forecast horizons. Imposing some parsimony on such models has huge payoffs. Force the data to really tell you something.

I think it's ok to say that on average, across lots of time, the model is a decent forecaster.** But...

Could it predict our huge recession?


No models I know of really predicted the Great Recession with any reasonable amount of warning (and yes, a few people seem to have seen it coming, but a lot of them have since been shown to be stopped-clock predictors). I have some thoughts on this problem later in this post.

Here are a few charts of the model trying to predict the Great Recession. Here's what the model forecasts when estimated on data through the 3rd quarter of 2008 (click for larger image):

So here's what's going on in the chart above. The solid black line plots the actual data for GDP. The black dashed line with + (plus) signs is what a random walk model (the benchmark!) predicted for the path of GDP. The rest of the lines show the path for different simulations of my model; I simulated the model 100,000 times, and the red lines surround 80 percent of the simulated paths.*** The blue lines surround 40 percent of the simulated paths. The black dashed line in the middle is the mean forecast. See how poorly it does? Let's see if the next quarter's forecast is any better.

Amazing! Forecasting from the 4th quarter of 2008 goes really well. But...

The model overshoots (actually undershoots), failing to catch this upward turning point just as it failed to catch the downward turning point (I don't show the unemployment forecasts, which are slightly better). This bad forecasting doesn't surprise me too much; I don't think many, if any, models are great for forecasting turning points. But there are other excuses people could make, probably along the lines of things not included in the model (like ad hoc policies).

For those who didn't read the previous section, this doesn't mean the model is an epic failure. It actually forecasts ok on average. But just when good forecasts are most needed, it doesn't perform (though it still beats some other models, and likely beats a lot of more informal forecasting methods).

The problem of forecasting

I've written before that I don't think the purpose of the study of economics is forecasting. The value of economics as a field comes primarily from its ability to provide understanding of economic phenomena and do good quantitative policy analysis. The majority of economists spend little or no time trying to forecast things. We don't have a crystal ball or tea leaves. I did this project to satisfy a time series class and because I thought it'd be fun to think a little harder about forecasting, but this is not my main research agenda.

But it would be nice if we could predict recessions and other economic crises, right? Of course. But there are a lot of things that make that kind of predicting hard to do.

One difficulty is the most obvious: the economy is complex. It has a lot of moving parts. It is subject to all sorts of shocks and vulnerabilities (and they don't follow nice 2-parameter distributions, by the way). It's just really hard to predict the future of complex systems that involve a lot of randomness. Luckily there seems to be some inertia to the macroeconomy, allowing for ok prediction during "normal times," but when we really need good predictions they're not so good. It is somewhat easier to look at the past or think with models to come up with an idea of how things work or how policies might affect economic outcomes or counterfactuals (my BVAR model can do some really cool ones), and that's why we spend most of our time doing that instead of trying to tell fortunes.

Another difficulty is the Efficient Market Hypothesis (actually this isn't only a difficulty; it can help too). Some versions of this are really controversial, but a weak version just says that prices already reflect the information that is available to the public, so you can't predict their movements unless you know something the markets don't. This idea has huge implications for investing and financial punditry--entire careers exist that wouldn't if people understood the EMH. It also has big implications for forecasters, namely, that what you see in markets is sometimes all there is. There isn't much other information you can grab to say a lot about where markets will be tomorrow. The upside of this is that markets can do some of our forecasting for us, but the downside is that forecasters who think they know more than the markets are likely to be proven wrong.

Finally, I think the biggest reason we can't very well predict recessions in particular is that we don't experience them often enough. This is a good thing, of course; but it makes recessions difficult to study. Basically, we've had about a dozen recessions since we started collecting halfway decent data. We've had about half that many since we started collecting good data. We've had five recessions since the Fed started really doing its job in the early 1980s. And during the time of good data, how many recessions were as deep as this last one?

It's really hard to study something that rarely happens. It's really hard to find reliable predictors of something that rarely happens. It's a small sample of data. And it's extra difficult since by now most economists think that not all recessions are caused by the same thing. If there are several things that can cause recessions, and we've only had several recessions during the period of good data, it should not surprise anyone that our predictions are not very reliable. Some people say that economics is a young science, and that's part of it. But I think the bigger problem is that good data collection is a relatively new practice, and we need a much bigger sample. The good news is that the sample is constantly being expanded, both over time and in terms of richer "drill-down data" that allow us to study macroeconomic questions with micro data (see here).

The economy grows most of the time, so it's hard to make a model reliably predict economic contraction. The good news is that we're reasonably good at knowing how the economy responds to recessionary forces. But still, we're not great at quantitative prediction at any long horizon.

Throwing our hands up and saying that forecasting is a waste of time is not the solution. Forecasting is absolutely necessary for economic decisionmaking; we can do it systematically and transparently, or we can rely heavily on our cognitive biases or idiot pundits or people trying to sell us gold. Over time, my money is on the systematic approach. The important thing is understanding the limits of our knowledge and fostering realistic expectations in the consumers of forecasts.

So I think this is a mixed bag. In "normal times", a model like my BVAR is likely to do a better forecasting job than most or all other approaches. This is just a first pass at it, so its performance can be improved further with a bit of work (which I may or may not do). But I don't think it's going to predict the next recession, and it won't predict the one after that either.

*This model started out as a replication of this working paper; I even received some very helpful guidance from Saeed Zaman (an author on that working paper). This is a 17-variable BVAR (almost the same 16 variables as the working paper model, plus I add real residential investment); the variables are motivated by the Fed's forecasting needs and medium-scale New Keynesian models (which, after all, are typically solved by linearizing into a VAR format). I employ a number of order-selection criteria and settle on a 2-lag model. The prior is similar to the famous Minnesota prior with a few modifications.

**To be transparent, though, a better test of useful forecasting ability requires the use of real-time data--that is, data as they were when first released, rather than the more accurate data that have been revised. Real forecasters have to rely on real-time data, not knowing revisions until after they had to make forecasts.

***The posterior distribution of the coefficient vector is multivariate t (that's the beauty of the Normal/inverse Wishart prior--it is a natural conjugate). There are many ways to get the predictive density, such as Gibbs sampling or numerical integration (since all needed distributions have closed forms); I go with numerical simulation instead. The simulation exercise requires getting 100,000 draws of the coefficient vector from the MVT and 100,000 draws of the model's disturbance term (actually 100,000 for each forecast horizon), which I specify as multivariate normal with covariance matrix chosen empirically. So a careful modeler can improve forecasts if they have an idea of what sort of shocks may be coming.

Wednesday, December 12, 2012

Should I short your startup?

I've been looking at some data on startups. For today, I'm just looking at one table that somewhat illustrates the uphill battle faced by startups (once again, I just call any new firm a startup; I've heard that's not the hip buzzword lingo the kids are using. Whatever.).

This forthcoming paper by Manju Puri and Rebecca Zarutskie is a fairly comprehensive analysis of startups and venture capital. Whatever one thinks of their research design, the paper provides a wealth of stylized facts about startups and venture capital. Its main advantage is that it builds on the Longitudinal Business Database, which includes every private nonfarm business establishment in the US. Take a look at Table VI; click for a larger image and spend a few minutes (I only include Panel A).

Click for larger image

I notice a few things.
  • Startups that don't get VC assistance often fail; about half will have failed by age 5 and two-thirds by age 10.
  • Startups that don't get VC assistance are unlikely to be acquired. Of course, for most startups getting acquired isn't the goal. But your Facebook friend who keeps talking about his Boston/Silicon Valley/Silicon Slopes/NYC startup is probably hoping for acquisition or something similar.
  • Startups that get VC assistance are much less likely to fail than those that don't. But even for VC-assisted startups, by age 3 the probability of failure is higher than the combined probability of getting acquired or having an IPO.
  • Almost one-third of VC-assisted firms fail within 10 years.

Nothing on this table is causal, of course. The table doesn't show that VC assistance causes lower failure rates. You'd expect VC-assisted firms to do better if only because venture capitalists are actively trying to invest in firms with good survival prospects. Read the paper if you want to think harder about causality.

Also, there is a useful distinction to be made between the guy who starts an oil change shop with no intentions of further growth and the guy who starts a tech company with hopes of conquering social media. That distinction is lost in this table, but other parts of the paper look at things like industry and firm growth.

The broader point this table suggests, though, is that lots of startups fail, particularly those that don't get assistance from venture capitalists.

Saturday, December 8, 2012

My five favorite books of 2012

Actually, I'm allowing myself to choose from the books I read in 2012 that were published in either 2011 or 2012. In order of preference:

1. Turing's Cathedral (George Dyson, 2012). I wrote a little review of it here; this is a book about the invention and early development of computers, particularly the one built at the Institute for Advanced Studies. This one easily takes the top of the list; anyone who is not fascinated by John Von Neumann is crazy (not to mention Nicholas Metropolis, Julian Bigelow, and Alan Turing). There is plenty for everyone here; it appealed to me as an economist but also due to my (embarrassingly amateur) interest in math, computer programming, and history (lots of great WWII and Cold War stuff). It also hits a lot of other topics ranging from biology to weather forecasting. Read this book.

2. The President's Club (Nancy Gibbs and Michael Duffy, 2012). This is a book about relationships between presidents--past, present, and future. For example, I did not know about the interesting relationship between Herbert Hoover and Harry Truman. Some readers may be surprised to learn that Nixon was a key adviser to not only Reagan but also Clinton, who said that the death of Nixon was like the death of his mother to him (being among the Nixon-obsessed, I liked this vignette). Like everything I read about him, this lowered my opinion of Eisenhower. The book is full of interesting stories and reminded me that these are actually decent human beings when they're not trying to get elected to something--the post-Presidents are often likable, reliable, and useful to current presidents, regardless of past fights or the partisan divide.

3. Bismarck: A Life (Jonathan Steinberg, 2011). See Henry Kissinger's review here. I thoroughly enjoyed this book. The narrative is engaging, and the characters are fascinating--Bismarck most of all. If you like European history, diplomatic history, biographies of incredible people, etc., you will like this book.

4. The Signal and the Noise (Nate Silver, 2012). Silver is probably overhyped by now, but it was nice watching him embarrass the pundit establishment. The book is very good, providing perspectives on forecasting from many fields including weather, sports (of course), economics, and many others. It is a very readable, nontechnical discussion, though it does ask readers to learn Bayes Rule (as everyone should). I will add the caveat that while his discussion of economic forecasting seems mostly fair to me, he did leave out a newer class of models (Bayesian DSGE) that seem to be improving forecasts. This book is a quick read, and I think most people with interest in even a few of the fields he covers would enjoy it.

5. Unintended Consequences (Edward Conard, 2012). Ignore the ridiculous subtitle; I reviewed this book here. I liked this book. It helped me get better economic intuition, though his discussion of monetary economics is wanting (and a few other topics can be sketchy, like his weird theory about inequality). Conard combines a lot of high-level business experience with a pretty strong academic economics bibliography. I wouldn't recommend it as the only book you ever read on economics, but I would recommend it as one important component of a popular press approach to the subject. And before you assume he's another conservative hack (my initial assumption), be aware that this is heavy on analysis and hits both American political factions pretty hard on some of their more ridiculous stances (e.g., progressives on taxation and conservatives on immigration).

Honorable mentions:

  • Grand Pursuit (Silvia Nasar, 2011). This may have made my top five if I could remember it better.
  • Hypocrites and Half-Wits (Donald Boudreaux, 2012). This one probably deserves a spot in the top five, but it's a very different kind of book. You must buy it. I reviewed it here.
  • Fairness and Freedom (David Hackett Fischer, 2012). I enjoyed this one because I have ties to New Zealand and read a lot of NZ history, but I'm puzzled to see it on so many other top books lists. The approach was too contrived--NZ and the US just don't make for great comparisons. The US is too big and diverse in terms not only of institutions but also people (including indigenous people). I would recommend this book to my friends with ties to NZ but probably not many others.

Sunday, December 2, 2012

Startups and the Great Recession

...and by "startups" I just mean new firms (I know some people prefer a narrower definition). Anyhow, here are some charts I made from BDS data (click for larger images).

Figure 1
Figure 1 plots the number of new firms by year since 1980; these are administrative data observed in March, so the 2010 observation tells you how many firms existed in March 2010 that did not exist in March 2009. As you can see, there was a staggering decline in startups prior to and during the Great Recession.

Figure 2
Figure 2 plots the number of startups as a percent of the total number of firms in the (private nonfarm) economy, showing that startups didn't just take a beating in absolute terms. That declining secular trend is also noteworthy.

Figure 3
Figure 3 plots jobs created by startups. Again, a pretty epic decline prior to the Great Recession.

Figure 4
Figure 4 plots the component of the overall job creation rate (using the DHS definition) accounted for by startups [edit: the figure is mislabeled]. So the overall story is that new firm creation and job flows therefrom were decimated in the years leading up to the Great Recession, both in absolute and relative terms. Some of this may be explained by the Moscarini model, but I think there may be other things going on as well (such as housing collateral).

Saturday, November 24, 2012

"Buy local" doesn't make economic sense

Today is "Small Business Saturday," which apparently is a day on which we are supposed to buy stuff from business establishments that are owned by people who reside "in our community." I have no problem with setting aside a day to applaud small business and entrepreneurship--a very worthwhile and commendable activity. But whatever "warm fuzzy" consequences this may have for shoppers, it doesn't carry a lot of economic logic. A few points:

  • A common argument is that buying locally keeps the money in the community, and this is somehow a good thing. Russ Roberts dispatches with this one here; basically, if we buy things locally simply because they are local and not because they make sense in terms of cost and quality, we are impoverishing ourselves and, ultimately, each other.
  • The "keep money local" argument is also a little silly because it relies on an arbitrary definition of "local." Local means close to me. If it's good to keep money close to me, why should I stop at the level of my town? Why not keep money in my neighborhood and only buy stuff from people who live on my street? But even that isn't as close-to-self as I can get. See where this is going? The logic behind "buy local" leads to the demand that we only buy stuff from people who live in our house. Or our bedroom. That's stupid. We don't enrich ourselves by "keeping money local." 
  • Rather, we enrich ourselves through trade--exchanging our own resources (which includes the product of our comparative advantage) for the resources of others at a price that makes sense to both buyer and seller (reflecting the costs of production and the benefits of consumption). Russ Roberts nails the trade concept here.
  • Buying local solely to buy local is inefficient and, therefore, wastes resources. This is bad for the environment and bad for the economy in general (that's right--your environmentally conscious friends who want you to be a locavore are probably harming the environment). Steve Landsburg explains this here.
  • Finally, there is a less-common argument about supporting small businesses that relies on old research about who creates jobs. Basically, there is a political conventional wisdom holding that small businesses are the most important business category for job creation; some people might see this as a reason to support small businesses. Better research has shown that this conventional wisdom is bogus.
If you get warm fuzzy feelings from shopping local, do it. Otherwise, stick to making consumption decisions according to the net value you get from the product you're buying. That will often mean buying local; when it doesn't, you can still feel good about encouraging non-wasteful use of resources.

Addendum: Discussion with some other people prompted me to add a caveat about costly information. It's possible that some local firms would provide better consumer value than chains (for example, in industries where economies of scale aren't substantial) but suffer from the fact that information is costly: brand recognition dominates when determining whether local firms are good. See here for evidence that this may matter. For these reasons, initiatives like "Small Business Saturday" may be worthwhile. In this respect, the catch phrase should be changed from "buy local" to "check Yelp first."

Friday, November 9, 2012

How conservatives should think about immigration

If conservatives were as serious about their preferred economic principles as they claim, they would embrace more immigration--and not just high-skill immigration. Here's why:

1. Freemarket principles: Labor is a market. It's like any other market. In most areas of the economy, conservatives see the value of preventing government interventions. When it comes to immigration, though, conservatives loudly support severe government interventions in the labor market. This particular intervention includes fences and armed border patrol agents and all sorts of other money drains. The benefits of nonintervention in other markets exist in labor too.

2. No more nanny statism: Current immigration law involves government telling businesses they cannot hire certain kinds of people. Why is that the government's business? In other areas, conservatives recognize that employment should be seen as a mutually voluntary transaction between employers and employees. Why do conservatives abandon this principle when they tell employers they cannot hire immigrants (and expect employers to enforce immigration law)?

3. Wealth creation: It is a simple truth that we do not make ourselves wealthier by producing things at higher costs than necessary. Preventing immigration--including low-skill immigration--obliges us to overpay for labor that immigrants could do for less. It is a waste of resources. It's like banning construction firms from using power tools. Hiring immigrants to do jobs for which they have a comparative advantage frees up resources for other uses, expanding our wealth. Conservatives should abandon their belief that only high-skilled immigration is good for the economy.

4. Most conservative excuses for opposing freer immigration either are bogus or can be fixed as part of immigration reform. See, for example, this discussion.

There are many other reasons to support looser immigration restrictions, but the ones I mentioned should appeal to market-minded conservatives.

Monday, November 5, 2012

Facts about jobs that may surprise you

A lot of attention is paid to net job flows in the economy--the net amount of jobs created in a given month, quarter, or year. Less attention is paid to gross job flows, but they tell a very interesting story about the US job market.

Gross job creation refers to employment gains at expanding and new establishments (plants or business locations); gross job destruction refers to employment losses at shrinking or closing establishments.* These numbers can tell us a story about job reallocation, or flows of jobs between business establishments (and between industries, states, or whatever).

To see why this is interesting, consider the following figure (click for larger image).**

Note that the chart records job flows in a given year as of March of that year; the Job Creation (blue) bar for 2010 refers to employment gains at expanding establishments between March 2009 and March 2010.

The depths of the Great Recession are captured by the bars for 2009 and 2010. Observe that in these years, businesses hired around 14 million new employees! In a private workforce of 110 to 120 million total employees, 14 million hires is significant. Many businesses opened or expanded during the depths of the recession. This probably surprises some people.

The problem, of course, is illustrated by the Job Destruction (red) bars. Between March of 2008 and March of 2009, shrinking or closing businesses destroyed almost 20 million jobs. That's almost one-fifth of the US economy.

Job recessions can be easily seen in this chart. When the blue line is taller than the red line, the economy has added jobs on net. When the opposite is true, national employment has declined.***

What should we learn from this chart? I can think of a few things:
  1. Even during recessions, the US economy is amazingly dynamic. Overall unemployment may be increasing, but many workers still move between businesses or from unemployment into employment. The majority of workers separated from their jobs can be absorbed by new or expanding businesses.
  2. Not all businesses are harmed by recessions. Some businesses expand significantly or are newly created during recessions, while other businesses shrink or close.
  3. Likewise, during periods of strong economic growth, many businesses are shrinking or closing. It's not always obvious what this means: businesses could be laying off workers while profits increase if productivity is increasing (but those workers can typically find jobs at expanding businesses). In other cases, though, businesses are shrinking or closing because they are unsuccessful--bad ideas, bad management, bad local conditions, or bad luck--even while the overall economy grows.
  4. The net jobs numbers hide a lot of labor market churning--reallocation and creative destruction.
  5. Discovering the nature of job flows is important. Are there strong net flows from some regions to others? From some industries to others? From small firms to large firms or vice versa? From young firms to old firms, or vice versa? How do flows relate to the business cycle? These questions matter for both policymaking and personal/business planning. There is a large economic literature on each of these.
  6. For economist readers, these data should illustrate some limitations of representative firm models.

*More precisely, these job flows definitions are discussed here and are defined as:
(Gross) job creation at time t equals employment gains summed over all business units that expand or start up between t-1 and t.
(Gross) job destruction at time t equals employment losses summed over all business units that contract or shut down between t-1 and t.
These measures are good for measuring flows between business establishments. They are useless for measuring within-establishment flows. 

**Data from Business Dynamics Statistics (BDS). These are aggregated from establishment-level administrative data and are subject to long release lags. 2010 is currently the most recent year available. Note that the BDS can be thought of roughly as the "population" for both the CES and JOLTS surveys, except that those surveys also include government.

***The difference between the blue and red bar for a given year should roughly equal the sum of the popular monthly private nonfarm payroll growth measure released by the BLS (summed over 12 months from March to February).

Monday, October 29, 2012

Hurricanes and modeling

Source: The Weather Channel
I recently had a conversation with a businessperson that got me thinking about models. We first discussed the election; he predicted a Romney win based on some things he's seen locally (he lives in an important swing state) and his own instincts. He sees a lot of Romney signs in neighborhoods, and most of the people he works with like Romney. I expressed skepticism, noting that most of the models I follow are predicting an Obama win, including in that particular state. Knowing of economists' fondness for models, he said something like, "yeah well I think you place too much trust in models." This reflected also a broader attitude that we often see in politically vocal bussinesspeople. They know what they need to know about the world because they have deep, hands-on experience with the nuts and bolts of how it works. How could I possibly know more about politics in his state than he can? After all, he lives there.

Later in the conversation, we were talking about the ongoing energy revolution in the United States. He has some expertise in both the chemical properties of several energy resources and the industrial uses to which they can be put. He made a suggestion about optimal trade policy with regards to energy exports; the suggestion itself is not as important as was his implicit definition of optimal policy. For him, an optimal policy is one that benefits the specific industries with which he is familiar. It was something more general than, but related to, the old sayings about how what's good for GM is good for the country, or what's good for the manufacturing industry is good for America, or whatever. This kind of producer-biased sentiment tends to drive the bulk of our national economic policy.

This conversation occurred against the backdrop of the approaching Hurricane Sandy. The hurricane has more or less followed the path predicted by most of the models employed by NOAA and other modelers. I wonder how many people would reject the models' predictions, preferring instead to rely on their own gut instinct and personal experience with wind and rain to forecast the path of the storm. Most of us would find that silly, even though we know that the hurricane models have been wrong before (or have had to make serious forecast revisions at the last minute). Most of us don't think having lived through lots of wind and rain qualifies a person to make hurricane predictions.

The huge benefit of models is that they allow us to summarize complexity in a useful way. They won't always be right. But if they're built in plausible ways, they can tell us a lot of things that we can't learn from simple experience. In the case of Sandy, the big payoff of the models was that they told us the storm would make a sharp left turn, hit the coast of the mid-Atlantic/Northeast, and run into independent weather patterns moving east across the continent--creating a sort of "perfect storm" which is more dangerous than the basic classification of the hurricane itself would suggest. The model doesn't do anything to reduce the strength of the storm, but its predictions probably save millions of dollars and hundreds of lives.

Of course, weather modeling is very advanced compared with other forms of predictive modeling. But I think it's still useful for showing the benefits of modeling in complex systems generally. Electoral outcomes are pretty complex--the Electoral College determines the outcome of the presidential race, and it depends on a lot of state-by-state factors. A good model incorporates not only economic fundamentals and polls but also the mechanisms through which those factors have driven outcomes in the past. I told my friend explicitly, "the models aren't always right, but they incorporate far more relevant information than your gut instinct." In fact, probably most of the information he is using to make his prediction is also employed in poll-based models, along with a lot of other information he isn't using.

The other conversation--the one about natural gas--also reminded me of the usefulness of models. To economists, optimal policy is rarely defined in terms of benefits to producers. The purpose of production is to consume. We usually define optimal policy in terms of a welfare measure that captures something like consumer surplus or happiness. It's hard--really hard--to determine the effects of a producer-side policy on that kind of welfare criterion. It's easy enough to see how a policy that helps a specific industry benefits firm owners in that industry. It's even reasonably easy to see how it benefits employees in that industry. But it's hard to see how it affects people who have nothing to do with that industry but nevertheless fall under the stewardship of the relevant policymakers. What will it do to wages? Returns to certain kinds of capital? The price of complements and substitutes to the outputs of the targeted industries? How will it affect markets for other energy resources? Conceptually, this isn't so different from the difficulty of knowing what will be the effects of jet streams or continental weather patterns on Sandy. It's hard to keep track of so many considerations without a model.

Models help us get outside our own experience. They help us see that opinion may be more diverse than the handful of yard signs we see on our commute might suggest. They help us see that policies have effects on more than just the specific fields in which we work. They help us see that hurricanes might not behave how our intuition says they behave.

Models help us identify some of the variety of mechanisms linking complex systems. Models are best used with good judgment, but in many cases judgment alone is unlikely to adequately account for the wide range of things that matter. I'm not suggesting that any model is better than no model. I'm not saying Nate Silver is as good at predicting elections as NOAA is at predicting hurricanes.* I'm certainly not saying that economic modeling is an exact science. I'm simply saying that it's foolish to write off the implications of tested models whenever they contradict our own opinions.

*In his new book, Silver devotes an entire chapter to weather forecasting. He notes that hurricane predictions "have become 350 percent more accurate in the past 25 years" (page 141) and that "the science of weather forecasting is a success story" (page 127).

Friday, September 21, 2012

Money: It's all made up!


This post will be more colloquial in style than most of my posts. Apologies in advance.

It's just made up money

The main thing that differentiates the Federal Reserve Bank from other banks is that it can create money from nothing. This bothers a lot of people. Mitt Romney is apparently one of those people.
We're just making it up. The Federal Reserve is taking it and saying, "Here, we're giving it." It's just made up money, and this does not augur well for our economic future. (Source)
The candidate seems troubled by the notion that the Fed buys Treasury securities by simply creating the money it needs to pay for them. Romney has exceptional business experience and probably a pretty good grasp of economics. He has two degrees from Harvard and has been a governor. But he doesn't get monetary economics. That's understandable--business experience doesn't necessarily teach a person a lot about monetary economics, which is a very macro-level thing and can be pretty counterintuitive. But I think the general public is capable of grasping this, so here goes.

The key point that Romney is missing is this: it's all made up! The money floating around the US economy was all conjured out of thin air at some point. This isn't unique to Fed purchases of Treasury debt; every time a central bank does anything, it is either creating money from nothing or magically eliminating it from the world.

A common response to this revelation--that it's all made up--is something like, that's what you get when you have "fiat" money instead of tying money to something which has "inherent value" (gold is a common suggestion). But guess what: A gold standard would still have the government creating money out of thin air!

Suppose we abolish the Fed and replace it with a gold standard. Our wise Republican leaders decide that the dollar should be worth 1/1500th of an ounce of gold (i.e., gold is worth $1500/oz). This isn't some free marketeer's paradise. The gold standard is implemented by having the government promise to buy or sell gold at $1500 per ounce in unlimited quantities. The government is fixing the price of gold rather than letting the market decide what it should be. What does this mean? Well, if I dig a hole and discover an ounce of new gold, I can take it to Uncle Sam and get $1500 for it. In simple terms, this means Uncle Sam will fire up the presses, print up a fresh batch of Benjamins, and hand them to me in exchange for my gold. Thanks, Uncle Sam.

So it's all made up! Uncle Sam conjured that money out of thin air! And by the way, the event I just described may cause inflation (yes, you can still get inflation under a gold standard).

Paul Ryan has said we need "honest money." I have no idea what that means, but I know that a gold standard, or whatever, isn't any more "honest" than our current policy.*

The key point is this: if the government is going to be in the business of controlling money--whether through a central bank which targets inflation, as most developed countries have, or through a central bank which sets a gold standard, or a silver standard, or a commodity-basket standard, or a Junior Bacon Cheeseburger standard--if the government is going to control money, it is going to be conjuring it out of thin air some of the time. Gandalf would have made a great central banker (but not Dumbledore**). The only alternative to this is free banking--getting government out of the currency business entirely and leaving money up to private actors. That's not going to happen any time soon, and indeed Romney has not advocated it.

It's all made up, folks. That's how central banking works.

But does it not "augur well for our economic future"? That depends on the stance of monetary policy. By that I mean, is money too tight, or is it too loose? Romney seems to be implying that it's too loose, and the Fed is making it looser, so bad things are going to happen. People fall into the trap of thinking money creation, or the current level of the money supply, or even the nominal interest rate, are indicators of the stance of monetary policy. In Mitt Romney's mind, the fact that the Fed has created trillions of dollars means that money must be too loose.

But the money supply is a bad indicator of the stance of monetary policy. If I told you that tomorrow Wendy's is going to make 1 million Junior Bacon Cheeseburgers, would you say that Wendy's is making too many or too few JBCs? I have no idea. I need to know how many people are going to want a JBC tomorrow. If only 100 people want one, Wendy's made too many. If 10 million people want one, Wendy's made too few (but if they're smart, they'll raise the price until only 1 million people want one).

Money is no different. You can't just look at the money supply, or the rate of growth of the money supply, and think you have measured the stance of monetary policy. The best way to measure the stance of monetary policy is to look at inflation (or expected inflation) and growth in nominal GDP (or expected growth in nominal GDP). As Scott Sumner frequently points out, lately money has been tighter than at any time since Herbert Hoover! And indeed, even after the Fed announced QE3, market indicators of inflation expectations did not indicate excessive monetary looseness.

Sure, all these asset purchases by the Fed may lead to somewhat higher inflation. But it's very unlikely that it will get so high or unmanageable that the economy will be at risk.

But they're monetizing the debt!

It's true that recent and current Fed actions have involved creating money to purchase Treasury securities, and other stuff. Romney, after talking to one expert and ignoring the broader consensus of economists, complains (same source as above):
The former head of Goldman Sachs, John Whitehead, was also the former head of the New York Federal Reserve. And I met with him... You know, we borrow this extra trillion a year, we wonder who's loaning us the trillion? The Chinese aren't loaning us anymore. The Russians are loaning it to us anymore. So who's giving us the trillion? And the answer is we're just making it up.
First of all, who cares if Whitehead used to work at the Fed? His opinion obviously does not represent the views of people who are, you know, actually running the Fed. Regardless, I'm not so sure about Romney's claim about who is willing to lend to us. Like everyone in politics, Romney dramatically overstates the importance of China for financing US deficits; of course, China matters a lot, but China only holds about $1.1 trillion out of the total $16 trillion national debt. But that has gone up over the last 5 years--Chinese appetite for US debt hasn't disappeared.

More importantly, should it bother us that the Fed buys lots of US debt? In a word, no. Not right now. Not when inflation is averaging below 2 percent. Setting aside concerns about availability of safe assets for collateral and risk aversion, if the Fed can buy US debt without causing lots of inflation and effectively forgive that debt (since the Fed returns all profits to the Treasury), why not do it? It's about as close as you can get to a free lunch. Reduce the burden on future generations while simultaneously stimulating the economy. What's not to like?

The common response--which Romney gives--is that once the Fed quits buying assets, interest rates will have to go up. Sure. They'll eventually go up. Markets don't seem to be counting on rates going up dramatically in the future, but maybe they're wrong. Regardless, why should we care? Who cares*** if eventually rates go back up after we were able to spend several years borrowing for almost nothing and having the Fed cancel a bunch of our debt?

When inflation is low and unemployment is high, it's a perfect time for the Fed to monetize some debt.

I don't mean to suggest that monetary policy has no costs or tradeoffs; when we get a real recovery I'll be as hawkish on inflation as anyone. I'm just saying that in the current environment, the benefits probably exceed the costs (in expectation).

*Someone might argue that since the bank notes created are redeemable for gold, they are somehow more honest, or cannot be said to have been created out of thin air or something. This is not relevant to my point as long as dollars have purchasing power; and it's fairly obvious that dollars can have purchasing power even without gold backing. If you don't believe me, let me know when I can stop by and take all your dollars off your hands. Also, if history is any guide, under a gold standard there's still no reason to believe your banknotes will always be redeemable for gold.
**Okay, so maybe he would. I'm not being entirely precise here.
***This is abstracting from elasticities of government spending to interest rates. I'm sure those matter, but I think they're second-order here.

Friday, September 7, 2012

What's the REAL unemployment rate?

Source: Wikimedia Commons

Today the jobs numbers for the month of August are being released. Inevitably, I will see people claiming that the headline unemployment number "is not the real unemployment rate." The real rate, according to them, is much higher; and they often suggest that authorities are intentionally feeding us a misleading number for political reasons. Are these skeptics right?

More precisely, what is the "real unemployment rate"? I'm going to suggest that there is no such thing, and I'm going to explain a few of the ways we measure labor market conditions. Each has benefits and limitations which are relevant to the discussion about the state of the economy. This will be a short guide for non-economists.

A lot hinges on how we want to define "unemployed." To different people and at different times, that word might refer to anyone without a job (in which case, happily retired people and stay-at-home spouses would be "unemployed"), anyone without a job who is actively seeking work, or anyone without a job who wants one. Perhaps it should include those who have part-time jobs but desire full-time work. It might even refer only to people who have been searching unsuccessfully for work for at least some amount of time or people who were fired or laid off from their job (i.e., they didn't leave voluntarily). Since there is no common definition of "unemployed" that satisfies everyone's needs, the Bureau of Labor Statistics (BLS) attempts to measure several versions of unemployment.

Each of these indicators is derived from a random sampling of about 60,000 households, conducted by the BLS and the Census Bureau, called the Current Population Survey (CPS). The CPS and its products are subject to response error, sampling error, and seasonal adjustment error (topics for another post, perhaps); so it's probably foolish to get excited about small movements in any of the unemployment rates I describe below as they may just be statistical noise.

Headline/official unemployment (U3)

The unemployment rate most often cited in the media is called the "civilian unemployment rate." It is defined as the number of people in the "labor force" who do not have work, expressed as a percentage of the total number of people in the labor force. The key restriction that bothers many people is that "labor force" is defined somewhat narrowly as civilian adults (16+) who are available for work and are either employed or have actively searched for a job within the past four weeks. People who have been temporarily laid off are also included. People who don't have jobs and are not looking for work are not included in this definition of "labor force," and this is why some people claim that the "real" unemployment rate is much higher.

But this number doesn't just exist to give politicians cover. It serves a useful purpose for those who are trying to diagnose problems in labor markets. Discouraged workers who have given up the job search aren't actually in the labor market, so if we want to know something about whether the labor market is clearing--whether it is efficiently matching job seekers with employers--then we need an unemployment measurement that is limited to people who are actually participating in the labor market. In fact, this is a decent way to sum up the headline unemployment rate: it attempts to measure how well the labor market is clearing.

There are additional reasons for caring about the headline unemployment rate. Unlike some other popular measures, we have been collecting the standard unemployment statistic in a reasonably consistent way since 1948. It's immensely helpful to be able to compare labor market clearing conditions across time. Finally, because of its usefulness for measuring labor market clearing, the standard unemployment rate has long been used by economists for finding helpful economic relationships which can inform policy.

The use of this rate in headlines and policy discussions is not new or unique to the political leadership of a certain party. This is the number we've used for decades--and it's compliant with international standards.

So when people complain that this is not the "real" unemployment rate, they're really just saying that they don't care about what this number measures. Often, these critics care less about labor market function than they do about broader economic misery.

Unemployed, marginally attached, and part time for economic reasons (U6)

This measure is the one that journalists, thinking they're very clever, like to discuss when asserting that the headline number is a smoke screen. But it's not a secret; it gets released by the BLS along with the headline number, and you can find time series data for U6 from a variety of websites (like FRED).

U6 is the broadest measure of employment problems in that it counts the most categories of people; to qualify as "unemployed" (more precisely, underutilized), a person must be unemployed (as defined in the headline rate), underemployed, or marginally attached. Underemployed people are those who have work but do not work as many hours as they would like. Marginally attached workers are those who want a job and have searched within the past year but have given up (and have not searched for work within the last month).

This number helps us get an idea of how much labor the US economy could supply in better economic conditions; it also gives us an idea of how many people are hurting economically because of bad labor market conditions. So it is useful for knowing just how far we are from satisfactory economic outcomes. In that respect, it is an important number.

However, like any statistic, U6 has its drawbacks. It is too broad to tell us much about immediate labor market clearing. It is also a new number--the BLS began collecting U6 in 1994--so we lack long time series for comparing business cycles over time. Finally, by using such a broad measure of underutilization, we lack the ability to pin down the different issues plaguing labor markets.

So the U6 measure does not fully satisfy our need for a good description of labor market conditions; but for those who want a single estimate of how many people are unsatisfied with the labor markets, this one might work.

More restrictive measures (U1 and U2)

There are two unemployment indicators that give rates lower than the headline number. U1 counts as underutilized only those who have been unemployed (by the headline definition) for at least 15 weeks. U2 counts as underutilized only those who are unemployed (by the headline definition) due to involuntary separation from their last job (i.e., those who quit their last job do not count). Since even booming economic times are characterized by a constant flow of people between jobs, these measures are useful for distinguishing between those who have good labor market prospects but are just between jobs temporarily (or voluntarily) and those who are actually in trouble.

Other broad measures (U4 and U5)

Finally, U4 and U5 bridge the gap between the headline number and the extremely broad U6 measure. U4 includes those who are unemployed by the standard definition along with discouraged workers--those who searched for work within the last year but gave up because they believe no jobs exist for them. U5 includes the standard unemployment definition and all marginally attached workers--those who have become discouraged for any reason. Again, these measures have useful purposes, helping researchers, policymakers, and commentators distinguish between reasons for job-search discouragement.

Six unemployment rates, seasonally adjusted (click for larger image)
"UNRATE" is the headline number (U3)

Final thoughts

There is no way to perfectly measure the state of the economy, and there is certainly no way to summarize it in one number. The BLS attempts to provide a broad and deep picture of the employment situation that allows us to drill down on the nature of labor market problems (and we can supplement these statistics with a variety of other data releases to get a clearer picture). The US statistical agencies are well respected worldwide for good reason. They exercise political independence and seek to approach data collection and summarization using best practice from statistical science. That's true even if some data releases contradict your ideological priors or local anecdotal evidence.

There is no "real" unemployment rate, nor are any of the indicators we have useless; what matters is choosing the right number for the right purpose.

Friday, August 10, 2012

Doctors and Economists: Why not just admit that it's complicated?

Image from Wikimedia Commons

My wife suffers from chronic migraines. Over the years, she has worked with a number of doctors to choose medication and diet routines to address the problem. None of it has worked very well. This morning, we met with a renowned and clearly very competent neurologist with whom we've worked before. He carefully reviewed past efforts and asked her how the latest suggestions had been working out. Upon hearing that we still do not have a solution, he listed possible options for moving forward, noting the potential side effects of each and carefully hedging about how well we can expect any one approach to work. We chose an option. I then had the following conversation with him.

Me: So would you say that headache treatment is at the research frontier?
Doc: I'm not sure I would say that.
Me: It seems like the approach is largely trial and error. Is the problem that this is a complicated phenomenon, and we just don't understand it very well?
Doc: No. It's not complicated. It's just that different things work for different people, and it takes time to figure out the right combination.
Me: And we don't have a reliable way to know in advance what will work.
Doc: Yeah.
Me: Do a lot of people end up finding something that works? [I wish I'd been more precise with this question]
Doc: A lot of people do. [I wish he'd been more precise with this answer]

So we know that different things work for different people, but we really don't know what are the observable predictors of what will work for a certain person. That sounds like the research frontier to me.

I was also intrigued by his unwillingness to simply say, "It's complicated, and as a profession we really don't know much about it." I tried to give him an out so he could say that at low cost. He didn't take it.

This is interesting to me because I think the public has an exaggerated sense of what doctors really know and what they can accomplish. And, as we recently learned, the public also has unrealistic expectations for what the discipline of economics can tell us, particularly in the realm of prediction but also when trying to diagnose problems and prescribe remedies. Both our diagnoses and our prescriptions have large standard errors--they are constructed with considerable uncertainty. In both cases, the public often become disillusioned when they see evidence of non-omniscience. I wonder how much the professions themselves are to blame.

I simply don't expect doctors to know everything. I am completely confident that they know far more than I do about medical issues, so I don't know why they're so hesitant to show me the frontier of their knowledge. I would be a better consumer of medical services if I had a better idea of where the frontier is. It wouldn't make me trust doctors less--it may even make me trust them more. Likewise, I'm confident that, in general, economists typically know more about economics than anyone else, but they don't know everything. Could we have prevented the recent widespread public disillusionment with economists if we were better at managing expectations?

I would add that economists have the added difficulty of doing their work in the middle of highly political issues.

Related: Did the financial crisis discredit macroeconomists?

Wednesday, July 25, 2012

You didn't build that: Both Romney and Obama are wrong

I know I'm late to this party, but I do have a brief comment on this "you didn't build that" charade.

First, it's worth noting that Romney and Republicans did take Obama's remarks out of context in a way which changed their meaning. The most common quote I've seen from the conservatives is, "If you've got a business, you didn't build that." Obviously this sounds like Obama is telling people who have their own business that they don't deserve credit for it--the government did the work for them. That would be offensive and inaccurate. Building a business requires both tremendous effort and a high tolerance for risk. Those who are able to do so deserve the overwhelming majority of the credit, even if luck and government play some role.

But that's not what Obama said. Here's the whole quote:
If you were successful, somebody along the line gave you some help. There was a great teacher somewhere in your life. Somebody helped to create this unbelievable American system that we have that allowed you to thrive. Somebody invested in roads and bridges. If you’ve got a business -- you didn’t build that. Somebody else made that happen. 

Even if you think Obama really believes that government built everyone's businesses, it's a stretch to think he'd say it in public. I think a reasonable reading of the full quote suggests that Obama is simply stating that if you have a business, you didn't build the "American system" that enabled your business to succeed. That's largely true. Business occurs within an environment of things like government-provided infrastructure, the education system, and a set of laws which sometimes are conducive to commerce (sometimes). Fair enough. Romney was disingenuous to claim that Obama thinks Henry Ford didn't build Ford, and all the rest.

But that doesn't let Obama off the hook. He made this statement as part of an argument for higher taxes on the wealthy. His apologists have cited things like roads and schools as examples of things businesses didn't build that justify higher taxes on certain people. But there are two problems with that argument. First, roads and schools and other government-provided amenities aren't the result of some benevolent government entity, providing for its people as a mother feeds a child. Those things are paid for by, well, people, including (and especially) rich people. And I don't hear many people on the right complaining about having to pay for roads and bridges (well, except bridges to nowhere).

Second, the higher taxes Obama wants are not even meant to pay for roads and schools. Current taxes are more than sufficient to finance those items. The taxes Obama wants are meant to pay for our massive warfare/welfare state. Here's a chart from the Center on Budget and Policy Priorities (click for larger image):


Observe that transportation infrastructure and education comprise 5 percent of the federal budget (indeed, state and local government are more relevant for those categories). The overwhelming majority of the federal budget--81 percent!--goes to defense and entitlements, the warfare/welfare state. If you're looking for the drivers of government's insatiable thirst for tax revenue, there they are. Estimate the tax burden needed to actually pay for our whole budget, then divide that number by 20. Now you have roughly enough revenue to pay for the roads and bridges and schools. You could double spending on those items and still be able to lower taxes if it weren't for the warfare/welfare state.

So, setting aside businesses designed to feed at the government trough, it's not clear that Obama's point about a government-created atmosphere helping businesses actually justifies higher taxes on high earners. In short, Obama is using the benefits of infrastructure to justify taxes to finance the ever-expanding warfare/welfare state. That's disingenuous.

Now if only we had candidates with plans to rein in the warfare/welfare state...

Monday, July 23, 2012

Did the financial crisis and recession discredit macroeconomists?

This post is a short essay I wrote for another purpose reviewing Ricardo Caballero's paper, "Macroeconomics after the crisis: Time to deal with the pretense-of-knowledge syndrome." It's not necessary that you read Caballero's paper to understand this post--but if this topic interests you, his essay is a nice starting point. 

Caballero joins others in taking stock of the field of macroeconomics. The paper is motivated by the crisis, but he rightly excuses the profession for failing to predict it. “Knowing [crisis-relevant] mechanisms is quite different from arguing that a severe crisis can be predicted.” He focuses more on understanding economic phenomena. His specific complaints are about overemphasis on the “core” of macroeconomics instead of the “periphery.”

For Caballero, the “core” of macroeconomics consists of researchers employing DSGE models. While noting that general equilibrium is important, he takes issue with these models and what he suggests is economists' excessive reliance on them. In contrast, researchers on the “periphery” focus on specific economic problems without attempting to fit their models into large general equilibrium structures. These simpler models can isolate specific mechanisms and are good sources of intuition, but Caballero notes that their resistance to broader application limits their usefulness.

Caballero avoids making specific suggestions about methodological changes or new approaches. He does advise that economists gain a greater appreciation for complexity, but he does not reject the DSGE framework or its use in policy analysis. His points are well taken. Hayek's admonitions about the pretense of knowledge are just as relevant and necessary today as they were when he made them. That said, however, it is not clear what Caballero is trying to achieve with his essay. Far from taking issue with his central argument, I believe that he has no central argument.

It is true that few macroeconomists provided useful, specific predictions of what would unfold starting in 2005 and continuing to the present day; but, as Caballero notes, prediction is not a reasonable goal of economic science. Understanding economic phenomena, including business cycles and financial crises, is the goal. Caballero provides no evidence that the discipline is lacking in tools for understanding what happened in recent years.

Complaints about DSGE models and rational expectations are widespread. While it is easy to see shortcomings in these approaches, it is not clear that their use rendered economists unable to understand crises. The amplification of financial shocks is easily understood with models suggested by Bernanke and Gertler (1996) or others. The persistence of sluggish output following a crisis is well documented by Reinhart and Rogoff (2009) and others, and the reasons for such sluggishness are not impossible to ascertain using existing models. The discipline's focus on DSGE modeling has not rendered us unable to think about how counterparty risk and uncertainty created demand for bailouts. The portrayal of economists drawing blanks on how to explain the crisis is inaccurate.

Caballero raises interesting questions but fails to demonstrate that we lack the tools to understand the recent crisis. Criticism of the field is most useful if it (a) reveals flaws in our ability to understand economics, or (b) provides new approaches which allow us to dispose of conventions which are unrealistic. Caballero does neither. If the goal is understanding rather than prediction, our commentary is better applied to convincing people outside the discipline to temper their expectations. In the meantime, sufficient incentives exist for individual researchers to improve the science at the margins, as they have been doing and will continue to do.


Bernanke, Ben and Mark Gertler. 1996. The financial accelerator and flight to quality. Review of Economics and Statistics 78 no. 1:1-15.

Reinhart, Carmen and Kenneth Rogoff. 2009. This Time is Different: Eight Centuries of Financial Folly. Princeton: Princeton University Press.

Friday, June 1, 2012

One-armed economists and Potomac Fever: Suggestions for the lay public

Or, public policy is hard.

The well-known story goes that Harry Truman asked for a one-armed economist because his economic advisers always gave "on the one hand,... but on the other hand..." answers to policy questions. The best economists still give these kinds of answers, but when I read the econ blogosphere I occasionally encounter Truman's preferred kind of experts: economists or econ enthusiasts using the discipline as if it were conducive to easy answers to policy questions. In this post, I'm going to describe the problem and offer some suggestions for non-economists consuming economic analysis. (If you don't have time to read the whole post, skip down to the bullet points).

The truth is, economics is complicated. Lacking full laboratories in which to conduct experiments, economists and other social scientists carefully apply what can only be called second-best (or worse) methods to imperfect data to tease out reliable assessments of how things work. Further, not only do policymakers face a set of difficult tradeoffs for every policy, but even when preferences over those tradeoffs can be agreed upon it is rarely obvious how policy levers should be pulled to accomplish desired outcomes. This is all complicated by the fact that social scientists “are people too” and therefore are vulnerable to the same biases which afflict everyone else (hopefully we are careful to put in place checks against those biases). In fact, my hunch is that a lot of the one-armed economists have motives other than good social science.

I like to think that we as a profession are generally doing a decent job, but it can’t be said for all of us.

Politicians don’t like two-handed economists. Politicians like clear policy recommendations, preferably ones which entail no costs or possibility of failure. They also like it when experts will lend “credibility” to their policy platforms. Therefore, economists who are willing to provide clear-cut answers to any question while making it all look easy are highly likely to acquire the attention of policymakers. The trouble is that these economists are probably drawing conclusions based on ideology or personal interests rather than careful application of economic science. When politicians hire such ideologically committed experts, they are defeating the purpose of hiring experts in the first place. Policy prescriptions delivered with complete certainty and ideological purity can be obtained on the cheap from cable news; an education in economics is not necessary.

I think this is a significant problem in policy debates. Part of the problem comes from the fact that economists typically use analytical tools which average people don’t understand. This can make it difficult to evaluate the credibility of economists’ arguments and even more difficult to draw conclusions about economic research reported in the media. Therefore, I’ve brainstormed a list of pointers for approaching the economics profession, and the media's version of it, in a policy context. I might need to revise them upon further reflection, but consider them a starting point.

  • Postpone acceptance of papers cited in isolation unless you are familiar with the relevant literature or have the analytical tools to evaluate the papers for yourself. Research on most policy questions is likely to have a history of contradictory findings before consensus is reached. This does not mean there are no policy questions for which a reasonable consensus exists; it simply means that loud voices pointing to economic research in defense of their views should be able to point to evidence of consensus, not a solitary, cherry-picked study. At the very least, they should be able to provide a discussion of competing studies and why they are unsatisfactory. Of course, this rule can be hard to follow for studies on topics with little previous coverage (e.g., studies of the recent recession) or real-time analysis.
  • Even when a consensus around the existence of some policy effect develops, it is not obvious what sort of policy prescriptions should follow. Be cautious about commentators who make the logical leap from well accepted research results to unqualified normative conclusions. That leap is impossible to make without some sort of welfare criterion. Competent, well-read researchers can agree about the findings of a strand of literature and disagree about the policy implications of those findings. As James Zuccollo explains, “few policies are unambiguously good or bad, so almost any economist’s commentary provides ammunition for both sides of the normative debate.”
  • There are precious few policy dilemmas with obvious solutions; and even for these, responsible economists are likely to deliver prescriptions with a complement of caveats and cautions. Steer clear of commentators who promise costless solutions unless they have a long record of carefully making their case (like Scott Sumner).
  • In more general terms, social scientists are most credible when they deliver analyses and recommendations with a large dose of humility. Economist Greg Mankiw writes that, among economics pundits, “certitude reflects bravado more often than true knowledge” (my italics).
  • Ideological dividing lines in the United States exist due to a variety of historical factors which are probably orthogonal to any rigorously founded worldview construction. As Brink Lindsey argues, “There’s no epistemologically sound reason why one’s opinions about, say, the effects of gun control should predict one’s opinion about whether humans have contributed to climate change or how well Mexican immigrants are assimilating.” Therefore, it’s very unlikely that a person with rigorous training in economics would arrive honestly at a set of political preferences which map seamlessly onto the platform of one or the other American political party. Be skeptical of commentators who are loud cheerleaders for mainstream political candidates.
  • Credible economists often do policy work, but they hold their noses while they are asked to be team players. Greg Mankiw is a good example of this: he has worked for several Republicans, but he hasn’t embraced their nonsense about magical tax cuts. He currently works for Mitt Romney, but you can bet that he’s not cheering for the parts of Romney’s fiscal “plan” which require fantastical supply side effects to avoid adding trillions to US debt. Be wary of economists who appear to be team players when they are not employed by candidates or politicians. Odds are they are pining for a policy job or are wearing ideological blinders.
  • Be cautious of commentators who give lip service to opposing views without actually giving air time to their strongest argument (i.e., they fail the ideological Turing test). There’s nothing objective about caricaturing the other guys’ argument then snidely dismissing it as idiotic. This is par for the course in political polemics, but most debates among academic economists are likely to be too substantive for that sort of treatment.
I am not discounting the value of economic research. Indeed, I think it is probably the best guide to policy we have. Don’t misinterpret my comments as implying that economic research is unreliable or that economists are incapable of agreeing on anything. I simply think readers should exercise caution against the possibility that economic research is being abused by pundits and even other economists for ideological or political ends.

Also, I am not suggesting that economists are only credible if they have no political preferences. Rather, I am suggesting that credible economists arrive at political preferences after careful consideration of tradeoffs, reflection on relevant theory, and evaluation of empirical evidence. Having gone through this process, these economists can make policy recommendations which adequately reflect both the process of arriving at a conclusion and the amount of uncertainty associated with the conclusion. If they display little or no uncertainty, write it off as “bravado” instead of “true knowledge,” and hire a new expert.

My personal experience has been that my training in economics has gradually made me less and less certain about what I thought I knew. I am puzzled when I see similar training having the opposite effect on others. My puzzlement is mitigated, though, when I see that those others’ surety is highly correlated with their ex ante political preferences. I think you see the point.

Hayek was right: “The curious task of economics is to demonstrate to men how little they really know about what they imagine they can design.” We are studying a subject about which perfect knowledge and, therefore, perfect policy recommendations are impossible. Economists peddling simple, obvious fixes to policy problems are more likely to be snake-oil salesmen than credible social scientists. With that in mind, beware of economists who self-assuredly cheer on the politicians with their grand designs.

See also Andrew Gelman's comments on economics journalism; Hayek's warnings.

Friday, May 4, 2012

A few thoughts on inequality and redistribution

The topic of economic inequality has become prominent in popular dialogue in recent months. This is a pretty difficult issue. I don't have any big conclusions about it, but I do have a few scattered thoughts which are specific to the often implicit demand for increased wealth redistribution as a necessary and effective response to inequality. Some of these can be thought of as questions that the pro-redistribution crowd needs to answer before we become too committed to contra-inequality policies. If you don't want to read the whole post, skip to point #5.

Figure 1: Median Income by Quintile (and Top 5 Percent), 1967-2008, all races
Click for larger image
Source (H-3 All Races)
  1. Measurement: The situation is somewhat more nuanced than most of the popular accounts have suggested. In particular, there is evidence that consumption inequality is less pronounced than income or wealth inequality, and there are good reasons to think that consumption matters more than income or wealth (some people have criticized the study on which that link is based, but the criticism is that it understates consumption of the affluent--not that it overstates consumption of the poor). To be sure, the evidence is mixed--but it's not immediately obvious that income-based measures of inequality are the appropriate way to motivate policy. In particular, any study which does not account for existing taxes and transfers should be immediately discarded for use in the debate over redistribution (though such studies may be useful for thinking about other contra-inequality policies). Further, wage data do not capture things like improvements in working conditions, which constitute an unmeasured increase in real compensation. And even our continued difficulties accurately measuring inflation have consequences for the inequality data.
  2. Fad-based policy: This topic comes and goes in public consciousness; but research into the causes and consequences of inequality, and appropriate policy responses (including redistributionary), has been going on for years in both macroeconomics and microeconomics. But the public dialogue is more sporadic, giving emphasis to inequality when some event draws our attention to it (like recession). It's important that we make long-term policies based on long-term problems, not public reaction to a cycle of attention to this or that topic. If we want to make good policy in this area, we should consult the literature rather than drawing conclusions based on someone's blog (yeah, I know).
  3. Cyclical vs. secular policy: A related reason for policy caution is our current position in the business cycle (the cycle of recession and growth). Recession-induced conditions will not be with us forever. Further, the recession and long road out of it seem to be causing a lot of people to forget all the economic growth of the last several centuries. Even in the depths of recession, majority of Americans were better off than they were even 15 years earlier. Excessive focus on the down side of the cycle is myopic. The consequences of cyclical economic activity should be treated with countercyclical policies. Permanent policy changes should be motivated by secular trends.
  4. Is anyone actually getting poorer? People getting poorer is very different from people getting richer at a decreasing or unequal rate. A lot of commentators seem to be suggesting that not only has the gap between rich and poor grown in recent decades, but the poor have actually gotten poorer (see, e.g., Joseph Stiglitz). This doesn't seem to be the case, even using income statistics. Look at Figure 1 above, which plots median income for all quintiles along with the top 5 percent. And there's some evidence that available data overstate the degree of stagnation for lower quintiles due to mobility (see Russ Roberts, though there is some evidence of poor mobility). Additionally, there's that pesky fact about the cost of living: the truth is that the real cost of almost everything we consume has fallen dramatically in the last few decades. A lot of people get money illusion when they think about this. Go watch this video by Steve Horwitz, and never complain about long-term increases in dollar prices again. This long-term decline in the real price of stuff means everyone's purchasing power has increased. Everyone's. The likelihood that few (if any) groups of people have actually gotten poorer is crucial to my next point (#5). In fact, in some cases, the way we generate stagnation among "poor" people is by continually redefining poverty; using quintiles to define poverty is unwise (for example, if we define "poor" to mean the bottom income quintile, then we'll never make any progress alleviating poverty because 20 percent of Americans will always be "poor"). This problem with distribution-based measures of poverty (rather than need- or consumption-based measures) leads to my main point:
  5. From need-based to ex post distribution-based: Even supposing the Left's narrative about inequality and what caused it is completely accurate, the data themselves do not imply any obvious policy prescriptions (particularly in regards to redistribution). We can only get policy prescriptions after we've established what we want the goal of redistribution to be. I think most people think of the goal as being need-based: we want everyone to have what they need in terms of nutrition, healthcare, and housing. We might have debates about some marginal needs, like transportation and post-secondary education. But in general, we have programs in place which meet those basic needs (food stamp programs, housing programs, medicaid, Pell grants, etc.). If those programs are failing to meet people's legitimate needs, we should fix them (but note that a change in or persistent inability of programs to meet needs is different from a continual redefinition of "need"). Regardless, that's a separate issue from economic inequality, which is completely irrelevant for need-based welfare policy. Using inequality to prompt policy responses suggests that the goal of redistribution is no longer based on need; it is based on distribution. That's a much harder goal to deal with, both in terms of justification and specification. Assuming we're meeting everyone's needs, what are the arguments for expanding our redistribution mechanisms to serve some goal regarding ex post distributions? And, supposing we're convinced that redistribution for redistribution's sake is an appropriate objective, it's still hard to articulate precise goals. It's much easier to be specific (or to at least outline key areas for debate) when our goal is need based. When it's distribution based, it's pretty hard to imagine persuasive arguments for any specific outcome metric. Are we aiming for a certain Gini coefficient? Why that number? Are we aiming for certain wealth or after-transfer income levels for each quintile? Why those numbers? And on the resource side, are we looking for some "fair share" measure of tax burden? In 2009, the top 1 percent of taxpayers earned 17 percent of US income and accounted for 37 percent of federal tax receipts (source). So as things now stand, a small handful of people are paying nearly half of income taxes. What is it about that number that is not "fair," and what number would be fair? I'm not saying the current level is too high--or too low. I'm just asking for someone to define "fair" in context and justify it. When we transition from a needs-based to an ex post distribution-based welfare policy, we open up a can of worms. The whole idea seems pretty hard to justify; and if you can justify it, it's still pretty hard to imagine reasonable quantitative targets. And once you have those targets, it's hard to imagine avoiding significant unintended consequences, for example, in terms of incentives faced by those who would otherwise be upwardly mobile through the quintiles. Whatever problems exist with mobility across quintiles are not solved, and are probably made worse, if we establish ex post distribution-based goals of redistributionary policies. That's Panodra's box. And I don't mean the music website.
  6. You can't redistribute income: Supposing we decide that more redistribution is in order, it's important that we understand the limits of what redistribution can achieve. Taking money out of Warren Buffett's savings account and handing it to poor people affects the poor person in more ways than one. The first-order effect is more money in their pocket. The second-order effects depend on the opportunity cost of that policy. Savings always equals investment. At the extreme end, the capital in Warren's savings account might have funded a new factory where the poor person would have gotten a job; this means handing that guy money now is like killing the golden goose to get the eggs. So the amount of income accruing to the rich which can be productively redistributed (meaning, it can serve the purposes we're claiming it will serve) is not as large as some people seem to think. As Scott Sumner says, "The income statistics simply don't mean what progressives think they mean--something like 'resources available for redistribution'. . . . If you combine wage and capital income in the same aggregate, you are counting the same resources twice." Read that whole post. The lowest-cost way to redistribute is to take the cash Buffett was going to spend on consumption of private yachts. The highest cost way is to take the cash which would have been saved (and therefore invested in production of something)--or to take the cash which Buffett would have given to the Gates Foundation. But that high-cost way tends to be the easiest from a policy perspective. I am inclined to think that redistribution is going to be a really inefficient way to address the inequality problem (perhaps a better avenue is proving and addressing Mark Thoma's marginal product argument).
  7. Some context: By both international and historical standards, most of the poorest Americans are wealthy. Our discussion of the gap between rich and poor in America is really just a discussion about the gap between the rich and the very rich (to a person in true poverty, our debate might sound like this). Whatever moral justification exists for intranational redistribution must apply to international redistribution as well. That is, to be morally consistent in our crusade against inequality, we would probably need to take every American's stuff--including the poorest Americans--and give it to someone in sub-Saharan Africa (of course, that's not a sustainable policy, but neither is ex post distribution-based intranational redistribution policy). (I know that there may be arguments for fighting inequality which don't rely on moral imperatives--I'm not talking about those). Our tacit acceptance of international inequality damages the justifications for focusing on intranational inequality. I long ago abandoned the delusion that I have the intellectual and philosophical firepower to sort through the issue of international inequality, but it's clear that intranational inequality simply pales in comparison.
These aren't dogmatic, ideological reactions to the debate. These are requests for better arguments. The people going broken-record on the inequality data aren't being forced to explain what their goal is. If their goal is a transition from need-based to distribution-based policy, they need to explain how they justify that--and they need to tell us what their target is so we can start discussing the consequences of the idea. It's hard to take the complaints very seriously when the complainers can't provide a solid justification for the connection between the data and their vaguely defined policy preferences, can't provide specific targets and policy mechanisms, can't justify those targets, and can't work through their ideas to see unintended consequences. In absence of that intellectual and conceptual framework, it's very easy for conservatives to write off the inequality debate as envy politics. Until we have this conversation and do some analysis, we shouldn't be blindly demanding this and that policy with uncertain connections to the supposed problems we think we want to solve.

I'm not suggesting that poverty and stagnant mobility in America aren't serious issues deserving of policy attention; but it's important that we avoid the Yes Minister trap: something must be done, this (more wealth transfers) is something, therefore this must be done.