Macro Roundup

Tyler Cowen posted a draft copy of a paper he's working on that looks at much of the same research we have ("Are Ideas Getting Harder to Find, Robert Gordon book/papers). In short he does think scientific progress is slowing down, heavily reliant on Bloom

"...TFP growth probably is the best contender for how to measure scientific progress. And overall TFP measures do show declines in the rate of innovativeness, expressed as a percentage of GDP.... Cowen’s The Complacent Class: “During the period from 1919 to 1948, TFP averaged well over 2 percent a year, meaning that new ideas were contributing economic value at that rate, and in some particular years, TFP came in at over 3 percent. That represents the arrival of new ideas at a relatively high rate. Later, from 1948 to 1973, this measure of innovation still tended to average about 2 percent. But in 1973, TFP declined dramatically, often coming in below 1 percentage point a year, and for the most part, TFP stayed low until 1995, with an average of about 0.5 percent. The mid-1990s to early 2000s were a new golden age for TFP, which again at times ran at 2 percent or more...TFP slowed down, however, and in the most recent years, since the financial crisis, TFP has averaged well below 1 percent growth a year, only slightly above its disappointing 1973 to 1995 range. As productivity researchers John Fernald and Bing Wang put it: “Three out of the last four decades have seen business-sector productivity growth near 1.5%.....Many metrics of scientific advance show strong absolute progress over time. For instance, the annual growth rate of scientists is now over 4 percent, compared to a world annual population increase of about 1.1 percent. That means the number of researchers is doubling about every sixteen years, though note the most rapid growth often is coming in countries which are not on the scientific frontier. Still, in North America scientists are growing at a rate of 1.6 percent a year, alongside 1.5 percent growth in the EU, and growth of 1.9 percent in Japan...Bunch and Hellemans compile a series of 8,583 important advances in science and technology across time. Jonathan Huebner takes the 7,198 of these that occurred between the Dark Ages and writing his paper in 2005, in order to chart their progress over time. This index correlates with Asimov’s index of innovators with a coefficient of 0.85.....Again, Huebner’s approach looks at the rate of innovation per capita, and not in total. His view is that technological progress is getting more difficult, or declining. The peak decade is the 1960s in their data, with a slowing and decline in positive rates since then. But an alternative interpretation is that prior to the mid 19th century, with its development of communication, literacy, and higher education institutions, civilization was not developing at its full underlying potential. That enabled a high rate of innovation per capita, since there was a lot of low-hanging fruit. Since then, we have hit or are pushing against various limits, whatever they may be.Our massive population increase should not be used to increase an ill-specified denominator, thereby making scientific progress appear less impressive, because in reality that progress is reaching a much greater number of people than ever before. Undenominated by population, scientific progress shows what is mostly a very positive story: a massive and rapid increase over time, accelerating with the industrial revolution....Several other papers judge that new ideas are indeed harder to find. Most prominently, four well-known economists --Nicholas Bloom, Charles I. Jones, John Van Reenen and Michael Webb -- have written a paper “Are Ideas Getting Harder to Find?” Basically their answer is yes. They focus on measuring how many researchers are required, over time, to produce a significant breakthrough in scientific ideas...their core result is that increasing numbers of researchers are required to generate a given amount of percentage improvement in various areas, for instance as mentioned chip speed, crop yields, or mortality improvements. This paper is perhaps the strongest single piece of evidence on behalf of the claim that scientific progress is slowing down.At the macro level, the paper even finds the startling result that “research productivity for the aggregate U.S. economy has declined by a factor of 41 since the 1930s, an average decrease of more than 5% per year.” This again follows from matching the growing number of researchers against other metrics of progress. In essence it means that per researcher productivity, by their measures, is being cut in half about every 13 years.While we do accept the core conclusions of this paper, we also see it as too pessimistic about the course of scientific progress...If you still doubt the basic result from Bloom,et.al., there is secondary circumstantial evidence pointing in the same direction of their belief in a research productivity slowdown. For instance, the age at first innovation seems to be increasing in the US, as the stock of existing knowledge grows and it takes longer to absorb it before getting to the frontier. Doctorates are taking longer to complete, co-authorship is increasing, inventors are taking out fewer patents per capita, and research teams are growing larger. Scientists cite more papers in their research, and they take longer to get their first peer-reviewed publication. There seems to be a possible decline in the quality of Nobel prizes - fewer are for fundamental breakthroughs - in at least some scientific fields, such as physics, chemistry, and medicine. Finally, there are fewer “renaissance” men and women making impacts on several fields. It seems also that large scientific teams are less likely to make disruptive breakthroughs than are small teams. All of those factors are consistent with per person or per researcher innovation becoming harder and the notion that science and indeed society more generally is undergoing a steady bureaucratization...."

New Summer's IRS enforcement could raise revenues by ~1T over a decade, just an FYI
"...On balance, the CBO estimate does not lead us to doubt that a substantial investment in the IRS has the potential to shrink the tax gap by more than $1 trillion — and potentially more — in the coming decade. Extrapolating from the CBO and Treasury data suggest that $1 of investment in the IRS leads to $11 of additional revenue, which suggest that a $1.1 trillion increase in tax collections — shrinking the tax gap by around 15 percent — could be achieved by increasing IRS outlays by $100 billion...."

Good charts from Mark Perry illustrating that the “hollowing out” of the middle class is driven by people “graduating up”

“…Bottom Line: As can be seen in the visualization above, America’s middle class did start largely disappearing in the 1970s, but it was because they were moving up to higher-income groups, not down into a lower-income category. And that movement was so significant that between 1967 and 2018, the share of American households earning incomes above $100,000 more than tripled, from 9.7% to 30.4%....”

Ed Conard

3. Washington Post. “More Americans go without health coverage despite strong economy, Census Bureau finds“

“Median household income today is right where it was in 1999. We’ve seen two decades with no progress for the middle class,” said University of Michigan economics professor Justin Wolfers. “The economy is producing more than before, but the gains aren’t being shared equally.”

4. CNN. “US income growth stalled in 2018, Census says“:

Median US household income stayed essentially flat in 2018 at $63,200, breaking a three-year streak of increases, according to data released by the Census Bureau on Tuesday. Median income in 2018 was not statistically different than in 2007 or in 1999, which was the high point. “It tells a pretty darn depressing story,” said Justin Wolfers, economics professor at the University of Michigan, of the report. “Two decades with no progress for the middle class.”

But a closer look at the Census Bureau data reveals there are several reasons that it’s maybe not such “a darn depressing story,” and might be a much more optimistic story of income gains for Americans than is being reported by the media. Here’s why:

1. Gains in Median Income for Households and Workers Was Better Than Reported. The Census Bureau report provides data in its press release and in Table A-1 and Figure 1 showing the following:

a. US median household income increased by only 0.9% in 2018 from the previous year and that increase was not statistically significant.

b. Median income for family households (65% of all households) increased by 1.2% and median income for nonfamily households (35% of households) increased by 2.4%, and both of those increases are statistically significant.

Since all Americans live in either a family household or a nonfamily household, the median income for all US households rose last year, and the increases were statistically significant. The only reason the median income for all households was stagnant has to be because of the changing mix and size of US households, a point I highlighted yesterday in this CD post.

c. The Census Bureau also reported that real median income increased for all workers by 3.4% last year and by 3.4% for full-time, year-round male workers and 3.3% for full-time year-round female workers; and all of those increases were statistically significant (see top chart above).

Bottom Line 1: Despite the media’s sober story of stagnation for Americans, all US households and all US workers experienced significant increases in their incomes last year, see top chart above for a summary.

2. Comparing 2018 Median Household Income to 1999 is Distorted by Declining Average Household Size. Between 1999 and 2018, it’s true that real median household income increased by only 2.7%. But adjusting for the ongoing decline in average household size, which fell to an all-time low last year of 2.52 persons, the increase in median household income per household member increased by 5.9% over that period, or more than twice the increase in unadjusted household income (see bottom chart). One way to overcome the evolving and changing nature and size of US households is to compare the real median household incomes over time for households with a fixed, constant number of earners. For households with one earner, their real median income increased by 7.6% between 1999 and 2018 (nearly 3 times the increase of 2.7% for all households) and the increase for two-earner households over that period was 12% (more than four times the increase for all households), see bottom chart above.

Bottom Line 2: Because of the declining size and changing composition of US households over time, comparisons of median household income in two different years (like 1999 and 2018) is an apples-to-oranges comparison and will understate the household income gains on a per-person basis over time, and significantly understate the income gains for a households that have a constant number of earners over time (e.g., one-earner and two-earner households).

Here’s how I explained the income share trends displayed above in a post on CD a year ago discussing last year’s Census report, updated with this year’s data.

Yes, the “middle-class is disappearing” as we hear all the time, but it’s because middle-income households in the US are gradually moving up to higher income groups, and not down into lower-income groups. In 1967, only 9.7% of US households (fewer than 1 in 10) earned $100,000 or more (in 2018 dollars). In 2018, more than 1 in 4 US households (30.4%) were in that high-income category, a new record high. In other words, over the last half-century, the share of US households earning incomes of $100,000 or more (in 2018 dollars) has more than tripled! At the same time, the share of middle-income households earning $35,000 to $100,000 (in 2018 dollars) has decreased over time, from more than half of US households in 1967 (53.8%) to less than half (only 41.7%) in 2018. Likewise, the share of low-income households earning $35,000 or less (in 2018 dollars) decreased from more than one-third of households in 1967 (36.4%) to below one-third of US households last year (27.9%), a new record low.

Bottom Line: As can be seen in the visualization above, America’s middle class did start largely disappearing in the 1970s, but it was because they were moving up to higher-income groups, not down into a lower-income category. And that movement was so significant that between 1967 and 2018, the share of American households earning incomes above $100,000 more than tripled, from 9.7% to 30.4%. Many prominent people like Paul Krugman and progressive politicians like Sen. Bernie Sanders and Sen. Elizabeth Warren claim that American’s middle class has been declining, disappearing, collapsing, losing ground, vanished, stagnated, etc. But the Census Bureau data on household income over time displayed above demonstrate conclusively that those assertions are incredibly and verifiably wrong.

Think about it for a moment and let it sink in — in 2018 nearly one out of three (and more than 39 million) US households had annual incomes of $100,000 or more. And the share of American households with that level of income has increased by more than three times since 1967! Then compare that picture of a prosperous America with millions of middle-class households moving up into higher income groups to the narratives we hear all the time that the American middle class is: losing ground, falling behind, collapsing, stagnating, disappearing, fill in the blank ___________.

Income gains for US households and workers are much more upbeat than what’s being reported by the media

Following the release on Tuesday of the Census Bureau’s annual report on Income and Poverty in the United States, the majority of media reports painted a pretty gloomy picture of stagnating household income for Americans, here’s a sample:

1. CBS News. “Americans’ income barely inching up despite economic growth“:

Americans’ household income is barely rising despite ongoing economic growth and low unemployment, a sign the typical family is failing to see significant gains from what has been a record-long expansion.

The median household had income of $63,179 in 2018, not statistically different from the 2017 median, the Census Bureau said. On an inflation-adjusted basis, Americans families are earning just 2.7% more than they did in 1999, when median household income stood at $61,526, or 2007, when the median household income was about $61,000. The data illustrate why 4 in 10 Americans sometimes face what economists call “material hardship,” struggling to pay for basic needs such as food and housing. While income has barely inched upwards during the past two decades, costs for essentials such as health care and housing has soared, pinching budgets for many Americans.

2. Wall Street Journal. “Median U.S. Household Income Showed No Growth in 2018“:

Mark Perry, "Three charts based on today’s Census report show that the US middle-class is shrinking…. because they’re moving up," American Enterprise Institute, September 10, 2019, https://www.aei.org/carpe-diem/three-charts-based-on-todays-census-report-show-that-the-us-middle-class-is-shrinking-because-theyre-moving-up/

Mark Perry, "Income gains for US households and workers are much more upbeat than what’s being reported by the media," American Enterprise Institute, September 13, 2019, https://www.aei.org/carpe-diem/income-gains-for-us-households-and-workers-are-much-more-upbeat-than-whats-being-reported-by-the-media/

Three charts based on today’s Census report show that the US middle-class is shrinking…. because they’re moving up