Bad data vs bad robots
“Lake Wobegon’s Law”: in US surveys a majority of respondents think their economic situation is looking good, but the country’s is looking grim. A recent paper by Harvard’s Martin Feldstein blames bad data: GDP fails to capture improving living standards and feeds a pessimistic bias in the media.
Economic data are not meant to measure happiness. I respect Bhutan’s focus on Gross National Happiness; but when a government says, ‘GDP is a flawed statistic, we should measure happiness’, be suspicious — they always say so when economic growth is slowing and tough reforms are needed. Gross Domestic Product tries to measure economic output. While flawed, GDP is a good approximation, and economic output matters: migration flows mostly from countries where per capita GDP is low to ones where it is higher — it goes together with better access to electricity, better health care, lower mortality. But Feldstein has a point.
GDP measures economic output, but does a poor job at it. Let’s start with the obvious: it does not measure goods and services we get for free. The classic example is “goods and services produced within the home”, like when I change a lightbulb… This becomes a more glaring omission in the digital age: none of what we read or stream for free on the internet is counted in GDP.
Quality matters. If I can buy a higher-performance laptop for the same price as last year, I am getting more value. Feldstein explains how GDP does a poor job at capturing quality improvements and the launch of new products — even with the “hedonic adjustments” devised by statisticians.
Are we better off than we think? Feldstein argues that we are — and that deep down we know it.
“When in doubt, blame the robots”. Yale’s Robert Shiller thinks the opposite: the media paint too rosy a picture with headlines of record-high stock valuations, while fear of a robot-dominated future makes people spend less and squirrel money away, causing economic stagnation and low interest rates.
Two top economists, two diverging views — this is economics at its best…but also underscores the uncertainty that in part derives by the poor quality of the data. Here is my take:
GDP figures hide important improvements in living standards. As Feldstein notes, we often hear that the real income of the median household has not risen at all in the last twenty years; but I doubt anyone sitting smack in the middle of the income distribution would give up smartphones and two decades of medical advances.
GDP is a worse measure of economic output than it used to be. Feldstein suggests this is the case, because (1) the share of services in private sector GDP has risen to 70%, from 50% in 1950, and mismeasurement issues are especially severe in services; (2) the digital economy is an important new source of non-measured services. I tend to agree.
But the statistical flaws of GDP cannot explain away the recent slowdown in economic growth and productivity. Most academic studies find that mis-measurement issues cannot account for the bulk of the productivity slowdown (I have reviewed the literature here in Section 4). Even if the actual pace of productivity growth is faster than the official statistics, the fact that measured productivity growth slowed by a factor of six suggests there is ample to scope to accelerate it .
Inflation might be even lower than we think — but that does not mean we still need easy monetary policy. If real GDP is higher, inflation is lower. Aside from free services, we do measure all economic value in nominal GDP; therefore, if we underestimate real GDP, we overestimate prices. This, however, does not mean we are in a deflation trap requiring extended monetary support and ultra-low interest rates. Nominal GDP growth has averaged close to 4% in the recovery. Assume that prices have been falling at 1% per year — deflation! That implies the economy has been growing at 5% per year in real terms, with the unemployment rate falling back to record lows. This would be a very strong economy with prices driven down by innovation and efficiency gains, and healthy improvements in real wages consistent with robust private consumption — not an economy on life support.
The economy is not that strong, but it’s still robust — and here I disagree with Shiller. Over the last three years, personal consumption expenditure has been going strong, averaging 2.9% per year, just under the 3.1% pace of 2001–2006. The savings rate is averaging under 6%: it has simply reversed the plunge of the bubble years. And households are still happy to put money in a stock market at record-high levels. Shiller’s story of consumers hunkering down and squirrelling money away in near-zero-return government bonds, fearful of a robot-dominated future, does not stand up to the data — flawed as the data might be.
The productivity slowdown is real, and we need more investment and better policies to address it. But economic data is poor, and should be taken with a pinch of salt. GDP data is especially shaky. Every three months, the preliminary quarterly GDP release sparks headlines and frantic activity in financial markets…but bears just a passing resemblance to the final figure two revisions later. We have better high-frequency data on economic activity and confidence; they give us a more reliable and timely picture, but we are still left with substantial uncertainty, feeding disagreement on the economic performance and outlook. This uncertainty leaves greater room for error in policy and business decisions.
Dream of better data. My geekiest dream is that as digital spreads to industry it will generate more and better data: we will get real time information on the actual use and performance of every piece of industrial equipment and the value generated, giving us a much better picture of the economy’s performance and its potential. Eventually…who knows…we might be able to delegate monetary policy to an Artificial Intelligence managing the flow of a non-deflationary version of bitcoin.
For now, though, all we have is fairly bad data. We should use common sense to compensate. Recognize that as the economy changes, some economic statistics become less meaningful and we work with a bigger margin of uncertainty. Be less dogmatic. Focus on what we know tends to work: a strong business environment, investing in innovation and education. Create more flexibility in economic institutions and business models, so we can react faster when needed. And when in doubt, blame the robots.
Let me know what you think, here or on Twitter. First, though, click here to prove you are not a bad robot.
