Biden by 20
Based on historical patterns going back to 1948, Biden should beat Trump by almost 20 points in the popular vote. Of course, if anyone could blow this, it would be Biden.
Back in 1996, when I was still doing Left Business Observer, I came across a 1993 paper by Andrew Gelman and Gary King, “Why are American Presidential Election Campaign Polls so Variable when Votes are so Predictable?” It cited research showing that despite all the volatility in the opinion polls during the campaign, the results were fairly easy to foresee months in advance based on some fairly simple models.
Inspired by that paper, I developed my own version of such a model—and, given my statistical skills, it had to be a very simple model. It had just two inputs: the president’s approval rating and the yearly growth in real after-tax income per capita (aka disposable personal income, or DPI), both measured in the second quarter of the year before the election. It all worked surprisingly well. I’ve updated it a few times over the years with subsequent elections, and just ran the numbers for 2020, which is where the prediction of a Biden landslide comes from.
A few more details. The model predicts the share of the popular vote that “should” be earned by the incumbent party (which may or may not be an incumbent person) and by the challenging party. What’s shown below is the difference between the two, actual and as predicted by the model.

While it was sometimes off on the margins—though, all things considered, it’s pretty good—it still predicted the correct result in 16 of the 18 cases. The only ones it got wrong were 1960 (in which there were suspicions, never proven, of vote fraud to Kennedy’s benefit and Nixon’s detriment) and 1968, when the wily Nixon got his revenge by beating the hapless Humphrey.
At the bottom of this page is the history of the inputs and results. Approval is average Gallup approval rating for the president from April through June of the election year (which isn’t necessarily the approval for the candidate: for example, the 1960 approval is Eisenhower’s in the spring of that year, not Nixon’s); DPI is the yearly growth in after-tax income per capita for the second quarter, from the national income accounts; and margin is the difference between the incumbent party’s share of the popular vote and the challenger’s. The equations are shown at the bottom. As I said, it’s a simple model.
For 2020, I’ve used Trump’s current Gallup approval rating and an estimate of an 8% decline in real DPI per capita from a year earlier. Since the New York Fed’s GDP tracker is predicting an almost 11% decline for the quarter so far, -8% is a conservative estimate. It shows Biden beating Trump by close to 20 points, or a 60/40 popular vote. At 0%, which is virtually impossible, it still would have Biden winning by over 4 points.
I should attach some consumer warnings here. The model predicts the popular vote, so it called Gore the winner in 2000 and Clinton in 2016 (which, if we had a sane electoral system, they would have been). And 2020 is a completely wacko political year, featuring a lifeless challenger to a mad incumbent in the midst of a pandemic-induced economic crisis. But the conclusion here is that no matter how things look now, if we have an election, it’s Biden’s to lose, a formulation that admittedly may inspire more doubt than confidence in the prediction.

[Apologies that the table is a graphic and not text; WordPress is not table-friendly.]
Fresh audio product
Just added to my radio archive (click on date for link):
April 16, 2020 Yanis Varoufakis talks about life under COVID-19, the economic crisis, vultures stripping Greece, and democratizing the EU (includes bonus audio clip of Jim Cramer recalling his Trotskyist past)
Fresh audio product
Just added to my radio archive (click on date for link):
April 9, 2020 Jeb Sprague on CV19 in Haiti and the DR • Rossana Rodríguez- Sánchez on CV19 in Chicago and Puerto Rico • Josh White on the new leader of the UK Labour Party, Kier Starmer
Miserable employment report
This morning the Bureau of Labor Statistics reported that 701,000 jobs disappeared in March. Economists had been expecting about a third that number. Hardest hit were bars and restaurants, accounting for 60% of the loss. Also hit hard: retail, temp work, and, shockingly, health care.
One reason job loss expectations were relatively low was that the survey of employers on which the count is based is done during the week containing the 12th—in this case, between March 8 and 14. (No one is expecting anything but a torrent of bad news in the coming weeks and months.) As the graph below shows, survey week came before the surge in applications for unemployment insurance from 282,000 in the week ending the 14th (survey week) to 3.3 million the following week and 6.6 million during the week ending the 28th. It also came before the wave of stay-in-place orders, which began on March 20. Within a week, 20 states and 4 cities issued such orders. (There’s a helpful timeline here.) It’s striking that employers began shedding workers ahead of the closures, not a good portent for the April numbers.

Expectations are that the unemployment rate, which rose 0.9 to 4.4%, will rise by at least 10 points and possibly 20 or more over the next month or two. The broad measure of unemployment, U-6, which accounts for discouraged workers (those who’ve given up the job search as hopeless but have looked in the past year) and people working part-time who’d like full-time work, rose 1.7 point to 8.7%. There is just no precedent for this rate of job loss.
The monthly surveys of households, on which the official unemployment rates are based, began in 1948, so we don’t have good stats for the slide into the Great Depression. We do have highly unofficial monthly estimates of the unemployment rate assembled by the predecessor of today’s Conference Board, available from the National Bureau of Economic Research. Those are graphed below. At the time of the great stock market crash, October 1929, the jobless rate was 2.3%. A year later it was 9.0%. It took over two years to break 20%, finally peaking at 25.6% in May 1933. By some forecasts we’ll be there before summer.

Fresh audio product
Just added to my radio archive (click on date for link):
April 2, 2020 Dania Rajendra of Athena on the walkouts at Amazon • Lauren Kaori Gurley on the walkouts at Whole Foods and Instacart • J.W. Mason on the World War II economic mobilization as a model for a Green New Deal
The hits keep coming
Goldman Sachs attracted a lot of attention with its forecast that US GDP will be off 34% in the second quarter of this year. That is a very big number. It’s three-and-a-half times the worst quarter in US economic history since quarterly GDP stats began in 1947. (That quarter, by the way, was the first of 1958, the onset of a sharp recession, which featured, among other things, an “Asian flu.”) Here’s a little perspective on that number.
That 34% figure is annualized, meaning it’s what the total decline would amount to if the quarter’s rate were sustained for a full year. A 34% annualized decline works out to a 9.9% decline for the quarter alone.* Big, but at least it’s not a third.
Unless you’re a connoisseur of these things, though, you probably don’t know that we never fully recovered from the 2008–2009 recession. That point is made in the graph below. The line marked “trend” is based on the 2.1% average growth rate from 1970 to 2007, the year just before the Great Recession hit. The “actual” line is, as the name suggests, reported GDP per capita. The Goldman Sachs estimate for the second quarter is marked with the dot. If something like that forecast comes to pass, we will have undone the entire 2009–2019 recover/expansion cycle in a matter of months.

Note how from 1970 to 2007, the actual line bounces around the trend, rising above it in expansions (peaking around 1990 and 2000, for example), and falling below in recessions (like 1975 and 1982). Actual never strayed far from the trend—until taking a sharp tumble in 2008 and 2009, from which it never really recovered. Since 2009, the growth rate has averaged 1.6%. Last year, which Trump touted as the greatest economy ever, it managed to get back to the pre-2008 average of 2.1%, an average that includes two deep recessions (1973–1975 and 1981–1982).
At the end of 2019, actual was 13% below trend. At the end of the 2008–2009 recession it was 9% below trend. Remarkably, despite a decade-long expansion, it fell further below trend in well over half the quarters since the Great Recession ended. The gap is now equal to $10,200 per person—a permanent loss of income, as economists say. That doesn’t translate literally into a loss of $10,000 in personal income; there are lot of other things in GDP, like investment. And gains in personal income have been concentrated in the upper brackets for several decades, so that doesn’t mean the average American is $10,000 poorer than they would be had the economy recovered normally after 2009. It does mean we have a lot less in the way material resources than we should. And it suggested there were serious pathologies underlying a superficial and often strange “prosperity.”
That’s all gone now. Regardless of the exact number, we have almost certainly entered a very sharp downturn, one that could rival or exceed that of the early 1930s, though at a much faster tempo. We could experience in months what took three or four years to unfold after the 1929 stock market crash.
Goldman is expecting a rapid recovery later in the year. I find that hard to believe. A shock like covid-19 isn’t easily recovered from. Even if we find our footing in two or three quarters, we’ll probably see another permanent income loss, unless we undergo some serious structural reforms.
Yes, GDP is a flawed measure of material well-being. It says nothing about what the economy produces, at what human and ecological cost, or how it’s distributed. But GDP is a useful shorthand for the principles around which our society is organized. This analysis helps explain why things have felt so unsatisfying despite cheerful economic headlines for the last five or seven years. And it’s only going to get worse, and probably a lot worse.
*Normally, you can annualize a quarterly rate by just multiplying by 4, or “quarterize” an annual rate by dividing by 4. Such approximations are close enough with the small percentages associated with the ups and downs of US GDP. When the numbers get large, however, that trick doesn’t work because of compounding. The formula to compute the real quarterly rate from the annual one is ((1+-0.34)^(1/4))-1, which yields -9.9. Or, if you want to annualize -9.9, it’s ((1+-9.9)^4)-1, which yields -0.34. For simplicity’s sake I’ve omitted the percent sign.
Fresh audio product
Just added to my radio archive (click on date for link):
March 26, 2020 James Meadway on the economic dimensions of the coronacrisis (article here) • David Quammen on zoonotic diseases like covid-19, which leap from animal to human and wreak havoc
Fresh audio product
Just added to my radio archive (click on date for link):
March 19, 2020 David Himmelstein of Physicians for a National Health Program and CUNY on how US health policy got us to this desperate pass • Helen Yaffe on Cuban interferon and COVID-19, and the country’s biotech industry and health system (YUP article here)
Why UI isn’t enough
I’m going to be posting a series of commentaries on the current crisis. Here’s a quick first
It’s odd to see Democrats like Pelosi and Schumer objecting to Republican schemes to send everyone a check for $1,000, maybe two. Of course, one- or two-off checks for $1,000 won’t pay many of the the bills for very long. But talk of means-testing right now looks mean, cheap, and politically suicidal.
Schumer says that rather than write checks, we should expand unemployment insurance (UI) benefits. It would have to be some expansion. Benefits are low, of short duration, and available to a smaller share of the unemployed than in the past.
Right now, the average UI check is $372 a week and the average duration of benefits is just under 15 weeks. That works out to a total of $5,515. While well above $0, it still won’t take you very far. During the worst months of the last crisis, in early 2010, the average check was $307 and the duration of benefits 20 weeks, for a total of $6,236. That’s about a tenth the average household’s yearly income ($63,179).
And the share of the unemployed drawing benefits has declined over the decades. Now, less than a third of the unemployed are drawing benefits. (Those are known as “continuing claims,” in the jargon). In the 1970s it was around 40%, sometimes as high as 50%. The unemployed include people who’ve quit their jobs voluntarily, or are just entering or reentering the workforce. If you take them out and compare continuing claims to the number of job losers among the unemployed, the numbers are higher, but still dispiriting: not quite two-thirds. It was actually lower in the aftermath of the 2008–2009 crisis, just over 50%. In the 1970s, it was between 90% and 110% (!). It’s all in the graph below.

One wonders what sort of expansion Schumer has in mind, but it would have to be a very serious expansion to be of serious help in the coming months. In the meanwhile, don’t complain about $1,000 checks.
Fresh audio product
Just added to my radio archive (click on date for link):
March 12, 2020 Kali Akuno on why black voters like Joe Biden • Dibyesh Anand on the belief system of India’s Hindu Fascists (book here)
Fresh audio product
Just added to my radio archive (click on date for link):
March 5, 2020 Andrew Bacevich, historian and president of the Quincy Institute, on the history and structure of the US permanent war mobilization (Harper’s article, The Age of Illusions) • Chris Brooks on the UAW bribery/embezzlement scandal (articles: ITT, Intercept)


