Government economic statistics are critical to understanding the economy and making policy. For example, the Federal Reserve relies on accurate information about inflation to make monetary policy. Accurate statistics are also important for understanding structural economic changes and economic performance.
Nowhere is this more important than in understanding what has happened to manufacturing output and business investment. For both, analysts and policymakers rely on the U.S. Bureau of Economic Analysis data. According to BEA data, things look pretty good on both fronts. Inflation-adjusted manufacturing output has been growing, leading most analysts and pundits to claim that all is well and that the massive loss of manufacturing jobs since the early 2000s has been the result of superior productivity growth, not declining international competitiveness and offshoring. Likewise, many argue that companies are investing in capital investment at a robust rate, supposedly refuting any claims, such as from American Compass’s Oren Cass, that our capital allocation and investment system is flawed.
The reality is that the BEA data is wrong, or at least highly misleading, painting a rosy picture when a dire one is more appropriate. It all boils down to how to measure the growth in output in the computer and electronics sector (NAICS 334) and investment in information processing equipment and software (IPES). To accurately track changes in the quantity of goods over time, statistical agencies must adjust for quality changes. This applies to output statistics and investment statistics alike. For example, a new tractor that is twice as powerful as an old tractor must represent increased output even if the price is the same. Statistical agencies use several different methods to adjust for quality changes, depending on the type of good.
Unfortunately, when these methods are applied to information technology (IT) goods such as computers, the rate of quality change is extremely rapid, and this leads to perverse and completely misleading statistics. Because of Moore’s Law (the trend of semiconductors and related equipment to double in speed every 18–24 months), IT prices have fallen dramatically while quality has increased rapidly. This means that, according to BEA, the computer and electronics industry’s output (NAICS 334), increased by 24,881% between 1980 and 2013. This one sector (accounting for around 7% of total manufacturing employment) saw such enormous improvements in processing speed that, according to BEA, it contributed 113% of the growth in manufacturing output between 2000 and 2011. In other words, it grew by more than the rest of U.S. manufacturing, which declined in output. This overstatement of output was so massive that it led to an overstatement of GDP growth by approximately 25% in the 2000s.
The United States is not producing 24,881% more computers than it was in 1980, and is likely producing significantly fewer because of offshoring. Instead, these enormous growth figures result from the statistical agencies’ quality change methods when they are applied to IT goods.
This misleading statistical adjustment method also vastly overstates investment in business capital. According to BEA, from 1980 to 2011, investment in information processing equipment and software (IPES) grew by 2,156%, over 31 times faster than the next fastest growing type of equipment, transportation, which grew by just 69%. According to BEA, between 1980 and 2011, over 85% of the growth in investment in private business fixed assets was accounted for by IPES. Between 2000 and 2011, that share rose to 96%. But this figure is also significantly overstated, painting a rosy and misleading picture of U.S. business investment. The typical bank, for example, did not increase its investment in computers by over 2,000% in the last 30 years. It might have gotten 2,000 times the processing power, but it didn’t in any real sense increase its capital investment to that degree. To put it plainly, these numbers are completely misleading and do not correspond to the real world.
Some will argue that these deflators are important because the IPES sector boosts productivity. But their impact on productivity in other sectors is already measured using existing BEA methods. This distortion relates only to the NAICS 334 industry output and IPES investment.
Economist Milton Friedman argued that economic models should be judged on the accuracy of their predictions rather than the validity of their assumptions. In the real world, computer and electronic products did not account for 113% of the growth in manufacturing output over the past decade. Likewise, IPES assets did not account for 96% of the growth in investment.
If the government had used appropriate, common sense measurement methods, policymakers and pundits would have known as early as the mid-2000s that U.S. manufacturing output was falling and that there was a crisis in capital investment.
If the Chinese government had manipulated our statistical system to fool policymakers into inaction and complacency, they couldn’t have done a better job than BEA has done inadvertently.
In the American Conservative, Oren Cass discusses how the American labor market’s failure to produce family-supporting jobs is fundamental to the nation’s problems.
A strong industrial base is vital to workers and their communities, the rate of technological and economic progress, and national security.
Daniel Moynihan once stated that “Everyone is entitled to his own opinion, but not his own facts.” This is no more true than with today’s debate over the health of U.S. manufacturing; a debate that is critical to get right if policy makers are to respond appropriately.