 On September 4, 2020, the Centers for Disease Control halted residential evictions in the United States for nonpayment of rent because of the COVID-19 pandemic. Eviction protection still needs to be the very top priority. Forcing thousands of people out of their homes during a pandemic will make a public health crisis worse. Many states and municipalities got there first, imposing at least partial eviction moratoriums starting in March and April. The theory behind these emergency orders was that if Americans were forced to leave their homes and move into more crowded settings, it would increase transmission of the coronavirus. The federal ban was supposed to expire at the end of 2020. Then it was extended a month, then two more months, then three more months, and then until the end of July. Many states and cities have also extended their moratoriums, in some cases through the end of September, even as COVID infection and death rates are plummeting. That's going to make matters even worse for landlords. The smaller mom and pop landlords who depend the most on rent for their personal income are being hardest hit by renters not paying. Tenants will end up picking up the tab. I pay it. I have to raise rents to cover the eight or nine tenants who aren't paying that I have. Small landlords who make up the majority of rental property owners in America have been forced to raise rents, remove units from the market, or refuse low income or credit impaired tenants. Were the eviction bans necessary to protect public health during the pandemic? Two studies that got widespread media attention, and that have also been cited by the federal government to support its policies, claim to show that the moratorium saved thousands of lives. These studies that have been widely publicized are seriously flawed. The underlying data are inconsistent and incomplete. The results are implausible. The size of the effect is wildly disproportionate to other public health interventions. And the researchers claim an absurd amount of certainty, despite the fact the underlying data are highly uncertain. Aaron Brown is a statistician who was taught at NYU and UCSD. He's also the former chief risk officer at the Hedge Fund, AQR Capital Management. If these studies were correct, they should be in big bold headlines of every newspaper in America. They would be the most dramatic public health finding in history. One study from Duke University researchers found that compared to a scenario without any eviction moratoriums, policies that limit evictions could have reduced COVID-19 infections by 14.2% and deaths by 40.7%. The second study, whose lead author is a postdoctoral researcher at UCLA, found that in states that had eviction bands but lifted them before the CDC moratorium kicked in, there were between 8,988 and 12,470 excess deaths caused by lifting the moratoriums. To have the CDC come out in the context of the pandemic and say that this was their rationale for the moratorium was really powerful. Brown says the first problem with these studies is that they draw strong conclusions from poor data. Information on where, when, and how many evictions occur in the U.S. is collected by a patchwork of local governments and courts and is notoriously inaccurate and incomplete. The important question for public policy is, if we have eviction moratoriums, will it reduce COVID deaths? In order to establish a causal link of that sort, you've got to work through it step by step. How do the eviction moratoriums affect evictions? How do evictions affect crowding? How does crowding affect virus transmission? How does virus transmission affect infections and deaths? The researchers don't even try to do this, and they can't because we don't have the data on each of those links. Brown says the study's final numbers are particularly implausible when compared to other public health interventions. The Duke study, for example, claims that had the federal eviction moratorium been in place between March and November of 2020, about 40% of all Americans who died from COVID-19 in that period would have been spared. In other words, about 122,000 lives would have been saved compared to a scenario with no moratoriums at all. To put that into context, the CDC estimates that 600,000 Americans have died from COVID-19 in total. The moratoriums might have prevented tens of thousands of evictions. Even if we ignore the offsetting effect of people moving into the vacated housing and reducing crowding, the eviction moratoriums might have reduced to COVID infections by a few hundred people. Even allowing for a few super spreader events, it's unlikely that the evictees would have caused more than a few thousand additional infections in the population using the case mortality ratios at the time gives us a number of approximately 100 deaths that might have been prevented by eviction moratoriums. But the researchers claim high confidence for thousands or tens of thousands of lives saved. That's just implausible. Statistical studies not only present findings, they also include what's known as confidence intervals to express the level of certainty we should have in the results. In both studies, the authors claim 95% confidence in their final conclusions. The most precise data both studies use is the number of COVID deaths, and serious researchers disagree about that by a factor of 60%. The studies rely on long chains of data, each link of which has much greater uncertainty than 60%. And the final uncertainty in their answer has to be the product of all the individual uncertainties. Yet these research claim they can evaluate the causal effect of eviction moratorium with standard errors of roughly 10% or 20%. And they have 95% confidence in those estimates. Mitten County's banned evictions now researchers from Duke say those places are seeing fewer cases and deaths due to COVID-19. The Duke study was co-authored by a team of academics with prestigious credentials, including a professor of economics with a PhD from Stanford and a quantitative epidemiologist and assistant professor of medicine. In the study, which was published in the National Bureau of Economic Research's working paper series, the authors didn't include the datasets they had used to reach their conclusions and listed just two of their sources of data. So reason contacted the authors and asked to see their math. As the US Department of Health and Human Services Office of Research Integrity states, once a researcher has published the results of an experiment, it is generally expected that all the information about that experiment, including the final data, should be freely available for other researchers to check and use. These guidelines apply to federally funded studies, but are standard practice among academic researchers. But Duke University denied our request on the grounds that the study hasn't been peer reviewed and therefore wasn't considered published work. The Duke paper is what's known as a pre-print, in which researchers make a study publicly available prior to submitting it to an academic journal. And yet the authors gave interviews in which they promoted their findings and the paper was widely cited in the press, appears twice in the Federal Register as part of rulemaking issued by the Consumer Financial Protection Bureau, and was cited in an amicus brief at the Fifth Circuit Court of Appeals to support legal arguments that eviction moratoriums have spared COVID-19 deaths. Duke also trumpeted the findings in a press release that encouraged members of the media interested in speaking with the authors to get in touch, though apparently not if that meant scrutinizing the actual dataset or code used to generate the paper's claims. The Duke researchers only cite sources for two of the data elements they use and those citations are incomplete. They claim to have gotten county level death data from the COVID tracking project, but that project doesn't have county level data at all. The Duke researchers also won't tell us when they accessed the data or how they matched it up with other data. If someone says they ran a multivariate regression, you want to know the exact code and adjustments they made. This is the only way independent researchers can validate the findings. Even without access to the researcher's data or code, Brown was able to discern a variety of ways in which the study misuses statistical techniques to arrive at faulty conclusions. First, the authors look only at differences within states when one locality had an eviction moratorium in place and another didn't, but they don't compare the COVID rates in entire states that had an eviction moratorium with those that didn't. Brown says it's clear that at the state level, those with moratoriums actually experienced more COVID deaths than states without them. The orange line shows states that always had an eviction moratorium in place. The black line represents states that never banned evictions. This finding, like the Duke study, is a correlation and almost certainly not causal. It doesn't prove that eviction moratoriums cause death. What it does do is put the burden of proof on anyone claiming a dramatic health benefit from eviction moratoriums to explain why it doesn't show up in the aggregate data. In fact, why we see the opposite. The Duke study is arguing that eviction moratoriums caused COVID death rates to decline and dramatically. If that were true, we should see the effect at all levels, both within states and between states. But the latter data doesn't support the author's conclusions, so they ignore it. The authors also ignore the timing of when people died from COVID-19. After states lifted their moratoriums, they had the same death rates as states that kept their bans in place and places that had moratoriums saw a major increase in COVID deaths in the summer of 2020, while in places that never banned evictions, death rates remained low. The other big data problem for the authors is that COVID death rates were higher when eviction moratoriums were in place than when they weren't. They don't make the obvious comparison of one place when it had an eviction moratorium versus when it didn't. All they do is they compare two different places, one of which had an eviction moratorium and one of which didn't. These two statistical manipulations allow the authors to get rid of the obvious objections to their study. All we can say is we have no idea from the data whether there's any causal effect one way or the other. Did eviction moratoriums increase COVID death rates, decrease them, or have no effect at all? We don't know. A new UCLA paper out this week directly links evictions to the spread of COVID-19. The study is a waiting peer review, but we wanted to make sure that policymakers had these numbers in hand as they're making decisions. Another widely cited study, which also hasn't been peer reviewed or appeared in an academic journal, looks at a period in which the CDC eviction moratorium wasn't in place during which evictions were partly allowed in 27 states. It argues that there were between 8,988 and 12,470 excess deaths caused by evictions in those states during this period. It also claims a 95% confidence interval for that finding. This study, written by an interdisciplinary team of researchers from several premier American universities, was cited by the CDC in its March 31 order extending the federal eviction moratorium. Like the Duke study, it was cited in the federal register and in the amicus brief to the 5th Circuit Court of Appeals. The study's lead author, postdoctoral researcher Catherine Lifehite of the UCLA Fielding School of Public Health, promoted the paper's conclusions in interviews with NPR, Vox, CNBC, ABC, and other outlets. Whenever you see numbers like 430,000 cases, 10,000 deaths, these are deaths that could have been prevented had the states maintained their moratoriums. It really is a human tragedy, not just numbers. But one reason reached out to her for an interview and said we were working with a statistician to examine the underlying data. She didn't reply to our query. This study is a model of transparent scientific reporting. The authors tell you exactly where they got their data and exactly what they did with it. I can duplicate their study. That also means I can pinpoint exactly what they did wrong. This study gets around the problem of the shoddy data on evictions by comparing death rates at times when moratoriums were in force versus times when they were not in force. Like the Duke study, it's looking to tease out the effect of a policy through correlation. The researchers note in some places when eviction moratoriums were listed COVID death rates went up. What they don't tell you is that in most places the opposite happened and in the aggregate the opposite happened. More than half the study's total of extra deaths they attribute to lifting eviction moratoriums come from Texas and South Carolina. If this were a genuine causal effect you'd expect to see it in every state or at least in most states and you'd expect to see it when the federal eviction ban were put in place. The fact that the authors can find a strong effect in only two out of 50 states shows this cannot possibly be a causal effect. Further complicating the study's results while it's true that in South Carolina and Texas there were no statewide eviction moratoriums during the study period major cities and counties within Texas had some type of ban and in South Carolina the state supreme court ordered strong restrictions on evictions statewide and nationwide many jurisdictions that didn't have explicit moratoriums still effectively halted evictions for months at a time as courts were shut down by social distancing directives. Others layered on emergency tenant protections subsidized housing costs or abated eviction proceedings for a period of weeks to months. All of these places are counted as having no eviction moratoriums at all. When Brown adjusts the authors analysis to account for local moratoriums and effective bans on eviction the authors effect goes away. Whether or not they did so knowingly Brown says the authors ended up manipulating the data to produce results that don't pass the sniff test. The elephant in the room as the last couple of decades have seen a floods of misconduct including fraudulent data no data and radically misrepresented data. We see it in the top institutions the top researchers top scientists. And the most discouraging part is institutions and journals circle the wagons to protect wrongdoers and slander anyone who questions the results. Duke had one of the most expensive and prominent examples of this when over a decade the university protected a medical researcher publishing false medical data. The university promised this would not happen again but the next chance they got when researchers were promoting dramatic public health claims but refusing to release their data the university closed ranks while at the same time pushing the findings out to any global reporter or policymaker they could find. The document Catherine Leifheit has compiled is dense. There are states, percentages, graphs and numbers lots of numbers. The study is still in pre-print that means it has not been peer reviewed yet. Normally we wouldn't cover a study like that. Thanks to journalists like you it is getting a fair amount of attention. Cancel the rent activists may have other reasons for wanting to make eviction bans permanent but they've seized on the results from these studies to pressure lawmakers and to push for a large expansion of the federal government's role in the housing market. An eviction crisis is looming. The moratorium is about to expire. Millions will be faced with evictions unless congress acts. I'm hearing numbers as many as 30 million people could be affected is that right? So that's one number that's been cited for sure. To the extent we're going to involve scientific research and policy discussions we should focus on the issues that have been studied by many people. Validated, sharing data argued back and forth. These shoddy one-off studies are just ammunition for people who want to put a link saying studies prove in an otherwise purely speculative article.