Kevin Drum

The Associated Press reports today on "The Great Grift," a combination of the plundering of COVID-19 aid programs by fraudsters and plain old mistakes in doling out the money:

All of it led to the greatest grift in U.S. history, with thieves plundering billions of dollars in federal COVID-19 relief aid intended to combat the worst pandemic in a century and to stabilize an economy in free fall.

An Associated Press analysis found that fraudsters potentially stole more than $280 billion in COVID-19 relief funding; another $123 billion was wasted or misspent.

....How could so much be stolen? Investigators and outside experts say the government, in seeking to quickly spend trillions in relief aid, conducted too little oversight during the pandemic’s early stages and instituted too few restrictions on applicants. In short, they say, the grift was just way too easy.

The story is long and a little hard to follow in places, so I simplified things with a summary chart:

Mistakes in handing out unemployment checks plus a small number of mistakes in mailing out stimulus checks add up to $123 billion. Total fraud depends on whose estimate of the Paycheck Protection Program you believe. The lower estimate produces a total of $182 billion in fraud. The higher estimate produces a total of $280 billion.

Add it all up and you get a grand total of $300-400 billion, or 7-10% of the pandemic money spent so far. A big part of the fraud came from the Economic Injury Disaster Loan and Paycheck Protection programs, both of which were administered by the Small Business Administration. SBA, which is a smallish agency, was overwhelmed by the vast scope of the programs and the need to get money out fast:

In the haste, guardrails to protect federal money were dropped. Prospective borrowers were allowed to “self-certify” that their loan applications were true. The CARES Act also barred SBA from looking at tax return transcripts that could have weeded out shady or undeserving applicants, a decision eventually reversed at the end of 2020.

Michael Horowitz, the pandemic watchdog chairman, criticized the government’s failure early on to use the “Do Not Pay” Treasury Department database, designed to keep government money from going to debarred contractors, fugitives, felons or people convicted of tax fraud. Those reviews, he said, could have been done quickly.

....In less than a few days, a week at most, Horowitz said, SBA might have discovered thousands of ineligible applicants. “24 hours? 48 hours? Would that really have upended the program?” Horowitz said. “I don’t think it would have. And it was data sitting there. It didn’t get checked.”

Unemployment fraud and mistakes were similar, but on a state level. Most states simply weren't prepared for a sudden, huge increase in unemployment payments, and their systems were often old and creaky. In Ohio, for example, State Auditor Keith Faber says nobody was prepared for what hit them:

The state’s unemployment agency “took controls down because on the one hand, they literally were drinking from a firehose,” Faber said. “They had a year’s worth of claims in a couple of weeks. The second part of the problem was the (federal government) directed them to get the money out the door as quickly as possible and worry less about security. They took that to heart. I think that was a mistake.”

There is, obviously, an inherent tension between getting aid out fast and getting it out right. They're both legitimate goals. Was a 7-10% penalty worth it in order to help lots of people quickly? I'm not sure I can confidently answer either way.

Eric Levitz makes an interesting point today about Donald Trump and his obsession with holding onto classified documents. The Department of Justice, far from indicting him for political reasons, actually went easy on him:

In January of last year, in response to a subpoena, Trump returned 197 classified documents to the federal government. Despite his willfully retaining those documents for months, the federal indictment released last week does not charge Trump in connection with any of them — which is to say, the DOJ gave Trump a pass on 197 potential counts of willful retention of national defense information. Instead, it charged him with only 31 counts corresponding with the number of highly classified documents Trump knowingly withheld from the government in January 2022 and the FBI later obtained.

Trump held onto hundred of classified documents for a year after leaving office, even after repeated requests for their return from the National Archives. He gave them back only after a subpoena. Even so, he did give them back, so DOJ let the whole thing slide.

But then Trump practically forced their hand. Even after NARA requests, even after a subpoena, he held on to dozens more classified documents and did everything he possibly could to hide them. The government got them back only after the FBI seized them in a search of Mar-a-Lago months later.

This is the key difference between Trump, Biden, Pence, and Hillary Clinton. Biden had a few classified documents in his possession after his vice presidency, but he promptly and voluntarily returned them when they turned up in a search. Ditto for Pence. Hillary never had any classified documents at all, merely some private email conversations that were mostly about things that were unclassified at the time but were later retroactively classified by the CIA in a routine feud with the State Department over classification standards. (The items that were classified at the time they were sent referred mostly to things that had been reported in the news.)

It's all simple, and even Republicans know it. Unlike anyone else, Trump actively and persistently held onto classified documents and conspired with others to keep them hidden. That's the difference. If he had returned them when the subpoena was issued, grudgingly or not, DOJ never would have done anything. But Trump refused, and at that point DOJ hardly had a choice. It was all but impossible not to indict him.

A few days ago we learned that US spy agencies knew in advance of plans by Ukraine to bomb the Nord Stream pipeline. Today there's more. We not only knew about it, we warned Ukraine not to do it:

The U.S. Central Intelligence Agency warned the Ukrainian government not to attack the Nord Stream gas pipelines last summer after it obtained detailed information about a Ukrainian plot to destroy a main energy connection between Russia and Europe, officials familiar with the exchange said.

The message, delivered by CIA officials in June, followed a tip the CIA received from the military intelligence service of the Netherlands, these officials said.

....The CIA then received information that Ukraine had called off the original plan, according to a U.S. official....At a meeting with a European counterpart in October [after the bombing], CIA Director William Burns told his counterpart that available evidence didn’t point to Russia. When asked if it was Ukraine, he said, “I hope not,” according to an official present at the meeting.

The CIA thought Ukraine had given up on the plan after the warning, but apparently they only delayed it. This was certainly ballsy behavior from a country that desperately needed billions of dollars in military aid from the US for its war with Russia.

Sunset in the French town of Vernon, about 40 miles downstream on the Seine from Paris.

Here's something from news of the weird:

Toyota’s latest engineering feat is an electric vehicle that behaves as if it has a stick shift, complete with revving sounds and faux gear shifts.

....The system generates fake sounds and simulated gear changes as well as genuine stalling if a rookie driver fumbles the controls, although the company is still debating whether to keep that last feature when it commercializes the technology, a spokeswoman said.

That's . . . interesting. Car companies barely sell any sticks on their conventional gasoline cars anymore, so why would EV drivers want one? “We want to be able to deliver a sense of ‘wow’ to customers,” explained Toyota's head of EV.

Whatever. But if Toyota's EV accelerates from 0-60 in 4 seconds, like a lot of EVs, even experienced stick shift drivers will have to be pretty fast on the clutch.

Bloomberg projects that peak oil usage is coming in six years:

Back in the day, many of us predicted that peak oil would arrive by 2015 or so. But that was based on the ability to produce oil, and it turned out to be wrong thanks to the shale oil revolution. Bloomberg's forecast, by contrast, is based on demand. They estimate that demand for oil will peak in a few years thanks mostly to the adoption of electric vehicles, while demand for petrochemicals and non-road transport keep rising.

As an aside, it turns out that all of us peak oil fans were mostly right back in 2005 when we predicted oil production would hit its peak around 2015.¹  The big question at the time was the future of OPEC oil production, and Saudi production in particular. And guess what? OPEC production has been flat (actually down slightly) ever since then, and Saudi production is up only slightly (about 9%), nowhere near the 15 million barrels per day maximum they claimed at the time. Total worldwide production of conventional oil has been flat since 2015, with all the growth coming from unconventional sources, primarily shale oil.

In other words, aside from shale oil and a couple of other small developments it turns out that we really did reach peak oil by about 2015.

¹Curious about what I said back then? It's all in a long, multi-part post I did for the Washington Monthly. Part 1. Part 2. Part 3. Part 4. Epilogue. Nut graf: "New discoveries will continue to decline and oil production will reach its peak in about 10 years."

Weekly blue-collar earnings were up 4% in May compared to April. However, earnings have been flat for the past year, and after their big jump at the start of the pandemic have been dropping steadily ever since:

Earnings for all workers fell 1% in May after adjusting for inflation.

The CPI inflation report for May was released this morning, and it's the same story as it's been for the past few months: headline inflation is down but core inflation is stubbornly steady:

On a trend basis, headline inflation is almost down to zero, but core inflation is around 5%, the same as it's been since October of last year. This is mostly because energy prices are down substantially, which affects headline inflation but not core.

However, the story is a little different if we take out housing, which has been artificially raising the core rate for many months:

Core CPI without housing has risen since October, but is down to 4.2% in May. This is likely a more accurate measure of core inflation.

I've long been puzzled by something related to remote work. There's a big gap between two different metrics:

• Based on a variety of estimates, the number of office workers who actually work in the office (i.e., not remotely) has declined from 90% to about 70% today.
• But every estimate of office occupancy in big cities suggests that it's declined from 95% to 50%.
  .
  

Why is the office occupancy rate so much worse than the actual work rate? One possibility is that the occupancy rate is for big cities, and remote work is more widespread among workers in big cities compared to smaller towns and cities.  Today I ran across the Stanford Work From Home project, which confirms that this is indeed the case:

These numbers are hugely overstated because of a flaw in WFH's methodology. But the flaw is consistent across the survey, so it's OK to compare remote work in big cities vs. non-cities. As you can see, there is a difference, but it's pretty small. It doesn't come close to accounting for the difference between in-office work, which has dropped 20 percentage points, and office occupancy, which has dropped 45 points.

So what's up? What am I missing?

A couple of days ago the Director of National Intelligence declassified a report written by an advisory group about the dangers of Commercially Available Information (CAI). This is information compiled by private companies and then sold to whoever pays for it—and it includes an astonishing array of data about individuals:

CAI can be obtained from public records, sometimes digitized from paper originals, such as information about real estate transactions that can be found in local title offices or courthouses. It can be obtained from smartphone and other software applications, often in the form of software development kits, that collect information from devices in the U.S. and abroad. And CAI can be obtained from cookies and other methods, sometimes associated with real-time bidding for sales of online advertising, that track end users as they browse the Internet.

....CAI can also be combined, or used with other non-CAI data, to reverse engineer identities or deanonymize various forms of information.

CAI is widely used within the intelligence community, and one of the advisors' biggest concerns is location data. If the federal government wanted to track you or me, it would normally have to show probable cause and get a warrant. But what if you can just buy tracking data instead?

The government would never have been permitted to compel billions of people to carry location tracking devices on their persons at all times, to log and track most of their social interactions, or to keep flawless records of all their reading habits. Yet smartphones, connected cars, web tracking technologies, the Internet of Things, and other innovations have had this effect without government participation. While the IC cannot willingly blind itself to this information, it must appreciate how unfettered access to CAI increases its power in ways that may exceed our constitutional traditions or other societal expectations.

CAI also implicates civil liberties. CAI can disclose, for example, the detailed movements and associations of individuals and groups, revealing political, religious, travel, and speech activities. CAI could be used, for example, to identify every person who attended a protest or rally based on their smartphone location or ad-tracking records. Civil liberties concerns such as these are examples of how large quantities of nominally “public” information can result in sensitive aggregations.

To summarize: (a) CAI is widely available, (b) it can be deanonymized so that it identifies specific individuals, (c) among many other things, it offers tracking data on individuals, and (d) no warrant is required to use it.

What to do? Should government agencies be allowed to use this? If not, does it make sense that literally anybody can use it except for government agencies?

In any case, this is not just theoretical. The Defense Intelligence Agency is already using commercial tracking data, although it segregates US location data into its own database with limited access. Approval for access is apparently rare,¹ but it notably requires only agency approval, not a warrant signed by a judge.

The best answer to this problem would probably be to prohibit commercial collection of location data in the first place. That's not likely to happen, nor is it likely that government agencies will be prohibited from buying commercially available databases. But the second or third best option is to at least require a warrant to use information that would otherwise require it. Unfortunately, Supreme Court precedent on this is fuzzy. It's something they should take up.

¹DIA says that permission has been granted five times in the past two-and-a-half years.