GDP and Other Mysteries



Octavian Report: The Leading Indicators examines the history, uses, and misuses of major economic statistics. How did you choose this topic?

Zachary Karabell: We’ve come to live in this world that is intimately shaped by a very limited set of numbers, and I’m always curious about how we come to live in the world that we’re in. And one thing that becomes apparent when you start asking, “How did this happen? How did it happen that we have defined our national success largely by reference to one number, namely GDP, gross domestic product. How did that happen?” — this wasn’t always the case. Lincoln didn’t get up and say, “Gross domestic product even in spite of the Civil War has increased and therefore I’m a good president.” And Franklin Roosevelt barely was able to say that.

The question is: do these numbers depict what is going on in the world around us with any degree of accuracy? When you start looking at the background and the history and then at the world we live in today, it becomes clear that while GDP and unemployment and trade figures and inflation numbers do provide some insight into how these things work systemically, they are a product of the moment in time when they were created. They were all invented in the 1930s and the 1940s and they were all invented largely to help the United States and Great Britain deal with the Great Depression and with World War II. That’s it. That was the world those numbers were designed to help those policymakers navigate.

The result is we’re really good at measuring mid-20th-century industrial nation-states, because that was the nature of economies then. We don’t live in that world anymore, but we use these numbers ever more in the world we are living in to navigate.

OR: Are they in your view generally directionally correct and accurate?

Karabell: These numbers measure what they measure, and they actually do a good job doing that. The problem is that what they measure and what’s going on in the world around us don’t line up so well.  Take unemployment: is the unemployment number directionally good at gauging whether employment writ large in any particular country is getting worse or better?

The answer is at best provisionally yes. As so many have been pointing out over the past couple of years, you have this somewhat unprecedented but definitely unfamiliar phenomenon of the unemployment rate in the United States going down along with the number of people who are actually working decreasing, the labor force participation rate.

As much as it would be great to answer the question of “Do these tell us something directionally that’s helpful,” the minute you start scratching the surface and you look at what goes into these numbers, that becomes a harder and harder question to answer conclusively.

OR: What do you make of the fact that people are investing or trading trillions of dollars based on these numbers?

Karabell: Using these numbers as trading hooks, as theses for forward action of any of the financial instruments that people trade I think is unbelievably questionable. It’s probably one of the reasons that most funds and most people who do this have seen results less than what they would have expected to produce for their investors or themselves.

There’s no way of really gaming out a GDP report as a direct one-on-one connection to earnings. GE or Honeywell or any number of large industrial companies are exposed 50 percent, 60 percent, and 70 percent to global trends rather than to U.S. trends. Does the U.S. really get to be the determinate direction of whether or not that company’s going to make its earnings? Does it matter to Amazon? GDP could be going down one percent and Amazon could get 10 percent more revenue just because they’re stealing a share, because they can afford thinner margins.

The idea that you can easily transform these economic statistics into viable trading theses, I think, is showing at best the law of diminishing returns and at worst leading you astray.

OR: The statistics themselves are revised over decades, correct? Including GDP?

Karabell: Right. And many of them are revised in subsequent months. The thing that provokes the trade is the initial release. Revisions, which happen frequently because almost all data when it comes out is provisional, more information comes in, better calculations, better adjustments — very rarely does the trade then get triggered by the subsequent revisions.

But there’s a market need and a kind of a public need for instant information, so we release GDP as quickly as possible. But the change disappears. There’s never a headline saying “Jobs — from two months ago, it was spectacularly wrong.” It’s always “X number of jobs was created this month.” And that is part of the media game of it. “They were created this month. They weren’t created this month.” They were a statistical adjustment or creation based on a limited number of sampling and a lot of math. That’s fine. That’s how statistics work. But you wouldn’t know that’s how statistics work given the public debate.

OR: I think you do a great job in The Leading Indicators, going back to the example of William the Conqueror’s Domesday Book, of giving examples of how flawed the statistics themselves are. Real policy decisions are made based on them; could you talk about your analysis of how the U.S.-China trade deficit is calculated?

Karabell: The whole relationship between the United States and China has unfolded really since China joined the WTO in 2001. It’s probably the primary bilateral economic relationship in the world. And most of the understanding of that relationship unfolds via these numbers, right? It’s an abstract statistical relationship based on a flow of goods and services.

The problem is, the way we measure that flow stems from the only way we know how to measure that flow. GDP and trade figures assume that you either make stuff within the boundaries of your country and then you sell it to your own citizens and then it boosts domestic GDP, or you make stuff and you sell it to another country, in which case it boosts your GDP, or you buy stuff from somewhere else, in which case you’ve sent your money to someone else. That sounds very facile, but that’s the entire framework of our current trade statistics.

The whole idea that something is made in one country and then either consumed in that country or sold to another country might have been true in the 1950s. Today, very few products — certainly very few manufactured goods and services — are actually made in any one country. And much of what China sends the U.S. that shows up as part of our $200- to $300-billion trade deficit, well, a lot of what we buy from China are products made by American companies sourcing their production to China. The iPhone, for instance, adds billions of dollars to the U.S. trade deficit with China every single year. Every time you buy an iPhone, every time you buy an iPad, it shows up statistically as two, three hundred bucks leaving the United States and going to China.

But the idea that those items were made in China in any fundamental sense is highly questionable, if not completely flawed. The OACD, the Asian Development Bank, and some individual economists have really started to look at this question. If something is assembled in China from parts that were made in five or six different countries with intellectual property that came from somewhere else, how much of the money that is ascribed to China for that object actually ends up in China? For the iPhone or the iPad, it’s probably less than $10 of a $600 or $700 retail product and a $200-plus wholesale product.

If only $10 is going to China, where is the rest of the money going? It’s going to Infineon, a German chipmaker. It’s going to Qualcomm, a U.S. chipmaker. It’s going to Samsung, which actually, even though Apple and Samsung fight, they still provide chips and property for each other’s devices. And it’s going to Apple, which provided the intellectual property and the marketing in the first place. You wouldn’t know any of that from the trade numbers because we have no way of breaking down tens of thousands of manufactured goods into their component parts.

Just to add to the complexity, not only does the trade deficit, in this respect at least, overstate the amount of money that’s leaving the U.S. and going to China and how much China is benefiting, there’s also the whole service economy to be accounted for. That doesn’t show up monthly. It does show up yearly, but it’s kept track of by a completely different organization within the government.

If a Chinese tour group comes to San Francisco, the way that gets calculated is as a U.S. export to China: the Chinese are coming here. If they come on an American carrier and they stay in a U.S. hotel and they buy stuff in the U.S., we have exported tourism services to China. This ought to offset some of that monthly trade deficit in goods. But those are kept as totally separate accounts, which you’d never know unless you were highly wonky.

OR: Do you think we actually have a trade deficit with China?

Karabell: Until we have a better way of doing this — and honestly, we’re not going to have a better way of doing this in the next five to 10 years; no one’s going to pay for this — we’re dealing with a framework that in my view is so inherently flawed given the world we’re actually living in as to be more harmful than useful.

OR: Early in your book you point out that there was a recent restatement of the way in which GDP is calculated.

Karabell: Yes.

OR: You point out that there was a major change: the equivalent of the size of the economy of Norway, you say.

Karabell: Yes.

OR: Can you talk about the concept of these statistics as moving targets and the methodology itself being revised over time?

Karabell: In the summer of 2013, the Bureau of Economic Analysis, which is the U.S. agency responsible for releasing and calculating GDP — and which has been revising its methodology for GDP constantly over the past 30 or 40 years — announced that, in fact, the U.S. economy was (statistically speaking) $400 billion bigger than it had been before, i.e. about three percent of GDP.

Now, it’s not as if everyone just got $400 billion richer. This was a statistical adjustment that said, “Just as we fail to calculate the intellectual property that Apple puts into its iPhones when we look at the trade deficit, we fail to capture the intellectual property development that companies and artists and all sorts of institutions do as part of our GDP.” They had been, until that summer, treated purely as expenses as opposed to consumption or investment. If you spent $10 million on R&D in a company to develop a drug or to develop a new device, that expenditure wasn’t counted toward GDP, it was just counted as a business expense.

But the money you spend to create that is like building a plant in the 1950s. Today’s version of building a factory is spending billions of dollars developing a drug or developing a new device or some information technology, so that the physical form that that investment takes isn’t a building or a factory, it’s an idea — an idea that then yields future return. And we ought to be counting that as part of our output, and that’s exactly what happened.

And then the Bureau of Economic Analysis went back to 1929 and did its best to revise all prior GDP to reflect this difference, so as not to make it look like GDP just bumped up three percent in the summer of 2013. Nor of course did that new number make any of us actually more wealthy. We were three percent wealthier than we thought, at least in terms of per capita income. But so what?

You didn’t know you had the money at the time, and the fact that you had it now didn’t increase your purchasing power or your spending. These are numbers, numbers created by human beings. They are measurements of a certain type of economy. As the economy has shifted, we’re scurrying to try to improve the way we look at it, and it will take a longer time to figure out how to measure it than it will to figure out how to live with it.

OR: You talk about in the book the underlying assumptions in data that the U.S. government in 2009 used for the stimulus — close to $1 trillion — based on certain numbers suspicious in terms of their accuracy. Can you tell that story?

Karabell: It’s fascinating that we’ve come to live in a world where the President of the United States, or of any country for that matter, can get up as President Obama did in February 2009 and make the following statement: “We’re going to spend $787 billion and this money will create or save 3.5 million jobs.”

Now, let’s take the partisan part out of it, because obviously that spending in that whole period of time has been the subject of huge partisan debate. What’s very interesting is the question of what allows for that level of precision such that the leader of a country can get up and say, “We’re going to spend X amount of money and it’s going to create Y amount of jobs.”

He didn’t get up and say, “We’re in the middle of a crisis just like in the Great Depression. Just like we did during the New Deal we need to take action to make sure that things don’t get worse. So we’re going to spend a lot money and we’re going to prevent things from getting worse and this will in fact make things better.” That’s a kind of descriptive statement. He didn’t say that. He said, “We’re going to spend this amount of money and it’s going to create precisely 3.5 million jobs.”

The only way you could say, “Well, if I spend X, I’ll increase Y,” is that you looked at your data in the 20th century and you saw that whenever government boosted its spending in order to do some sort of stimulus that the effect 18 to 24 months after that time was this percentage of jobs materialized. The problem is, if we’ve only been calculating GDP since at best the late 1930s and really not even then, and if we’ve only been calculating employment numbers at best since the late 1930s and really since the 1940s, then your playbook is 60 or 70 years of information. And it’s not like these stimulative efforts happen every year — so it’s not even 60 or 70 times.

The idea that these are mechanistic systems, that we have enough information and enough data to make that kind of statement with that kind of certainty should on the face of it be ludicrous. No statistician or scientist would say, “If you’ve got a hundred data points, you can make conclusions that are hard and fast.” They would say you can make those conclusions statistically, but your margin of error, the probability that you’re going to be wrong, gets higher and higher and higher the less data you actually have.

So we should be much more circumspect about what we think we know about future outcomes of spending. Again, this is true for economists, it’s true for Republicans, it’s true for Democrats. It’s true for everybody in that we don’t know enough. We don’t have enough information to make the kind of hard and fast conclusions we make about future outcomes. We shouldn’t be surprised when those future outcomes so rarely unfold the way we think they’re going to unfold.

OR: How accurate from a political standpoint do you think these numbers are in the United States and — perhaps more importantly — overseas?

Karabell: Every country in the world now measures GDP and measures national income and measures trade and inflation according to a globally determined set of standards largely articulated by the United Nations and then enforced — that’s a loose term — by the International Monetary Fund and the World Bank and other international financial organizations. That has led to every country in the world using a similar methodology. But the question is, even if the methodology is standard, are the inputs equally good or consistent? To some degree, we just don’t know. We know that every now and then, countries radically improve or revise what they’ve done before.

In early 2014, for example, Nigeria revises its national income methodology so that its GDP suddenly becomes bigger than that of South Africa. That may in fact be true. But it was true before it was statistically true if it is true. It wasn’t like Nigeria suddenly grew 15 or five times more quickly. Same thing happens with the U.S. and China, right? In May of 2014, the World Bank said, “We may not have been calculating this thing called purchasing power parity correctly. We may have been underestimating in a dollar-to-dollar sense and a cost-of-living sense the size of China’s economy.” And that therefore China’s economy may actually be bigger than the U.S. economy in 2014 — rather than in 2019, when they thought it would be. But just because you’ve revised your methodology about how far a dollar goes in China versus how far it goes in Chicago doesn’t mean that China suddenly got bigger more quickly. It’s a way of measuring two things that are incredibly complicated. We don’t fully know how to do it. We hardly know how to do it between Mobile and Massachusetts.

There’s this constant struggle to make the numbers consistent, and it’s less about actual corruption or actual fraud or actual misreporting. In many cases, it’s much more about, “Can these numbers capture some sort of underlying consistent reality?” There is fraud. The Argentine government under the Kirchners fired the statistical office in charge of publishing official inflation figures because they kept publishing higher figures than the Kirchners wanted. They actually then tried to form an NGO in Buenos Aires to do the due diligence themselves and the Ministry of Finance revoked their nonprofit status. Then Argentina was told by the IMF that if it continued to politically manipulate data in that way, it would be suspended or evicted from the IMF. Which would have been the only time any country ever did that, and lo and behold, the Argentine inflation bureau was allowed to function with a little more transparency and consistency and did indeed show higher inflation numbers.

Most countries in the world don’t go that far. Yes, China has regional misreporting of numbers, but at the end of the day the Chinese don’t have a real interest in misreporting their own GDP. Officials might, but the overall government doesn’t. They have an interest in making sure their economy is growing quickly enough to meet the needs of 1.3 billion people. I don’t think that it’s the problem of the data per se, I think it’s the limitations of the statistics.

OR: What do you think investors should take away from your book?

Karabell: I think there are very few questions investors will find are meaningfully answered by using these numbers as lodestars or as guideposts. Even the belief that with a certain amount of growth, of economic growth (i.e. GDP growth), there will inevitably be more employment, higher wages, inflation, and higher interest rates.

All those are a lattice of numbers that come together in a neat thesis, a thesis that clearly pertained for 60 or 70 years in the 20th century. 60 or 70 years of human society does not represent the entire spectrum of probable or possible outcomes for all these various systems and forces. That doesn’t mean that that thesis won’t happen. I’m articulating one that I know is out there, which is: growth up, monetary policy will tighten, wages will increase because the share of labor will demand more wages, and then you’ll have prices going up and interest rates going up.

It’s a neat thesis, and it pertained for a lot of the 20th century. The belief that that therefore will be the narrative for the 21st, just because these numbers represent a pattern, and which I think a lot of investors simply assume — I’ve heard so many times people say, “Well, this is what happens. This is what happens because this is what happened.” Every investor in the world, if they sell anything publicly, has to say, “Past performance is not indicative of future results.” We should have disclaimers on all of our data and all of our assumptions about future outcomes of this data: “Past patterns do not guarantee future results.”

But we don’t do that. We just assume, “Oh, this is a hard and fast pattern.” I think that’s worked so many times, and certainly in the past five or six years, so many people who rely on these patterns playing out have been scratching their heads. You could line up a hundred investors and I’m sure you would find 80 or 90 of them have been chronically perplexed by why interest rates haven’t gone up. Or why stocks have done so well. Or why corporate profits haven’t come down. You can argue, “Oh, it’s an anomaly, it’s an anomaly based on the financial crisis of 2008, 2009, and the mean, or this average, or this norm will reassert itself.” And that might be true. I’m simply suggesting, given the changes in the world that we actually live in and the limitations of our numbers to capture them, that the assumption we’ll revert to some sort of mean that only existed for 60 or 70 years based on a limited amount of data and a limited amount of time could be a colossal mistake.

There’s a lot of other information that we can all obtain about how companies are doing, about their fundamentals, about trade between nations. We have more data than ever before and we have an easier time finding it and we have more tools to manipulate it and analyze it than ever before. We can all do a better job, I think, utilizing that data. And we should all do a better job not cleaving to simple patterns based on simple numbers that I think are likely to lead us astray.