One of the major fronts of controversy during the second Trump administration has been, somewhat surprisingly, data: Elon Musk’s phony “efficiency watchdog” DOGE has been granted improper access to the Treasury Department’s payments data, as well as seeking access to healthcare and labor information, looking at Social Security databases, and allegedly entering the Internal Revenue Service (IRS) to snoop the tax data of American citizens and businesses. This is all at the center of various lawsuits against USDS (the full legal name of the DOGE agency) and is part of Elon Musk and Donald Trump’s attack on US democratic institutions. 

A second, somewhat less reported front in the War On/For Information is the disappearance of multiple government websites as well as data sources: mass retractions of papers onequality and diversity, as well as gutting the research and evaluation arms of major agencies and just generalized staff cuts at research organizations. Beyond questionable pushes into data-generating agencies, suppressing important information on pressing topics (for instance, bird flu), and generalized takedowns and blackouts of government websites, this widespread campaign of censorship opens a very disturbing possibility: will either DOGE or the Trump Administration’s lackeys interfere with American government statistics?

From the lying mouths to your lying ears

Statistical forgery is usually associated with autocratic and illiberal governments, as well as economically unsuccessful and extremely corrupt administrations. For instance, less democratic countries are more likely to fudge their official statistics. Comparing GDP growth to night-time electricity usage (which is highly correlated to GDP), researchers find that non-democracies tend to overstate economic growth by approximately 35%. The decision to lie on government statistics appears to be highly strategic and well-planned: African countries tend to misreport statistics to foreign donors in order to juice donations, and at the same time national governments are misled by subnational ones in order to juice grants and transfers. Other major offenders, at least notable ones, are Greece in the run-up to the Financial Crisis, Argentina from 2007 to 2015, and China basically since always. Likewise, countries vulnerable to balance-of-payments issues are also significantly more likely to report numerically suspect BoP statistics, seemingly in order to project strength and prevent the situation from spiraling.

American statistics, it should be noted, may have flaws of various kinds, but are generally accurate: the Billion Prices Project, tracking the online price of a billion goods in 2015 and 2016, resulted in remarkably similar readings to the official CPI. So with a good starting point, and since autocracy, economic failure, and corruption are all on the menu for the Trump and Musk duumvirate, it’s worth taking a close look at how they could pull it off. 

The first, and most obvious giveaway of statistical manipulation is that there are always whistle-blowers. In Argentina’s case, Secretary of Commerce Guillermo Moreno fired all the top technical staff of the top statistics agency in order to reduce inflation figures (from an estimated 490% to just 160%)—but many of those fired staffers produced their own “accurate” indexes, and were fined upwards of $125,000 dollars for disseminating “false information”. Likewise, Greece’s one-time top statistician Andreas Georgiou spent more than a decade being dragged through the Greek financial system for his revisions of government figures, which had purposefully underestimated the country’s debt burden and fiscal deficit in order to remain in the EU’s good graces. For whatever it’s worth, this appears to be a common practice in Europe, though mostly done via accounting trickery. So if former statisticians are fired en masse, especially at the top ranks, and raise alarms after which they are immediately charged with overwrought crimes, it’s definitely a sign. 

A second sign is that economic indicators do not correspond with each other, at least not following any reasonable explanation. Looking at 2021 vaccination statistics in La Matanza (Argentina’s fourth largest county), only around 60% of the population was vaccinated, compared to 80% to 90% in nearby counties. Similarly, the town had the lowest COVID fatality rate of its area, even controlling for age and income. Despite being one of the poorest parts of the country, La Matanza had as low a welfare takeup rate as San Isidro, the richest municipality in the Buenos Aires metro area. It also had both the lowest school enrollment rate of the metropolitan area and the lowest share of voters to population—meaning that there was no plausible age distribution that could generate these statistics. The likeliest explanation was that the 2010 census was manipulated—and the 2022 national census found an extremely sharp discrepancy in growth rates (from 3.9% yearly in 2001-2010 to 0.2% in 2010-2022) that isn’t matched in any other Buenos Aires metro locale. 

The point here is fairly straightforward: different data points, especially from different sources, have to tell a cohesive story that is also coherent with established macroeconomic relationships. China’s case is perhaps the strongest example: by many indicators, economic growth statistics are just too stable relative to other trends (which may extend to other key data such as imports), and as the country developed, its official statistics reported implausibly small changes in consumption patterns. While GDP statistics are closely watched, electricity-related indicators are less so, and the two can sometimes diverge quite strongly. Other indicators that match growth trends (known as “coincident indicators”) tend to match up well outside of periods of economic upheaval, where they de-sync for seemingly unexplainable reasons. 

Generally speaking, there is no smoking gun proving that China is fudging its statistics (for example, the numerical violations typical of hand-adjusted data are not present), but there is a wide shadow of doubt—which may even point to an understatement of growth. The biggest issue is in how China collects its stats: in a highly decentralized manner, with provinces reporting their figures to the national authorities and the authorities adjusting and collating them. This has led to a number of discrepancies, mostly in the industrial sector, generating large and persistent gaps between the two reported figures, and Chinese investment data yields multiple highly contradictory totals. The proximate cause seems to be the lower level authorities, which add “water content” (exaggeration) into their figures—for example, a number of Chinese cities have straight up admitted to forging their statistics. In order to reduce the “water content”, the government bypasses lower level authorities and accesses the data directly—with limited ability to ensure that the data remains high quality, and without removing the incentives to produce fake data altogether. 

The third way to tell that  government statistics were tampered is that… people just don’t buy them. There is an adverse impact on the quality of political choices, but it seems to be limited, and is mediated by partisanship and intellectual overconfidence—meaning that, at best, Musk could fool the stupidest, most partisan, and most overconfident people in the US and next to nobody else. Likewise, China’s changes to basically all its major statistical releases to cover up an ongoing economic slowdown fooled precisely zero serious professional analysts. And when households in Argentina were faced with inaccurate inflation figures, they only used some of the information: if phony inflation increased, they raised their expectations, but not if it decreased—so inflation figures were weighed by their reliability. This generally tracks with other research showing that inflation expectations are mostly focused on people’s price-related experiences, a loop that only strengthens if inflation is lower—so people will index on the supermarket more heavily than on, well, the consumer price index. This is usually not a problem because market prices tend to match official figures, but when they don’t, alternative measurements come out—subnational price indices (state or local CPIs), privately constructed statistics, or those published by other government institutions (Argentina’s Congress and the Central Bank both had their own CPI).

Prensa y garrote

Official data is extremely useful, especially if it is highly granular, highly detailed, and highly longitudinal, like the US government’s is. Private businesses and researchers make frequent use to answer weirdly specific and particular queries, such as “how much more did eggs cost in the Midwest than the South in October 2005” (the answer: 0.014 dollars). Economic statistics, thus, provide useful, important information for the public in general: inflation, employment, output, and quality of life data help drive major decisions such as when to go to school and what to study, where to live, and who to vote for. 

In a world where narratives overwhelm hard data pretty consistently, it was only a matter of time for the data to not just take the back seat but be molded to fit the narrative—as the Nobel Prize in Economics Laureate Ronald Coase said in an oft-paraphrased quote, “if you torture the data long enough, nature will confess”. But since fake government data is, as mentioned above, really bad at making people change their minds, what other purpose could actually be behind tampering and manipulation with official data? 

Well, let’s look at the effect: introducing fake information may not convince individuals on the aggregate, but it does generate a large degree of confusion—people just can’t figure out what the actual figures are. Media coverage also tends to muddy the statistics, since it is usually geared towards the negative—especially if the media is of opposite partisanship as the government. Confusion is a powerful tool for a government, especially an authoritarian one: it prevents the opposition from organizing around clear signals. This is especially true as political repression has moved from violence to information: dictators pay for media support much more than for outright bloodshed. In the words of Spanish dictator Francisco Franco, prensa (press) beats garrote (the club). 

As explained in The Power Broker, one of the major ways by which city planner Robert Moses evaded accountability for his actions was to not just avoid releasing the data on his proceedings (which, under laws that he had written, was kept under lock and key at various public authorities), but to only publicly announce fake, flattering statistics—while his office claimed that only a couple of thousand of families were displaced by his “great works”, biographer Robert Caro finds that the minimum plausible estimate is 500,000—that is, one out of every 28 New Yorkers at the time of writing (1974). Similarly, many headlines lauded the success of El Salvador president Nayib Bukele on the crime reduction front, without noting that soon after his regime took over, it modified the definition of “murder” to exclude a large number of victims from the official government tally. 

So the objective is not necessarily to get people to believe your numbers, but for them to not have any real or reliable numbers to talk about—that is, to control the flow of information and use “the data tap” to produce a deformed sphere of public discourse that is only molded by the desired conclusions Trump/Musk and their camarilla of lackeys. In this sense, nature will confess to whatever the torturers want.  

Hijacking the sausage factory

So could a “statistics coup” work in the United States? It’s hard to say. The American federal statistical system is, unlike basically every other country (except, well, China), highly decentralized and scattered across 13 statistics-generating agencies as part of either independent agencies (the CDC, Social Security and IRS statistics), as well as Cabinet-dependent organizations (the Bureau of Labor Statistics or BLS at the Department of Labor, the Census Bureau and the Bureau of Economic Analysis at the Department of Commerce, etc.) and other organizations such as the Congressional Budget Office and the Federal Reserve that conduct their own, independent statistical work. This is done either through work in other government databases (notably, IRS statistics), through direct work by the agencies (the BLS conducts surveys and interviews directly), or by coordination with state-level agencies and organizations. US states don’t appear to have much infrastructure in terms of their own independent data generation (though, in many cases, are involved in surveying and producing the data), but many important repositories of statistics are privately generated and/or handled, such as the University of Michigan’s Survey of Consumers, the University of Chicago’s General Social Survey, or the University of Minnesota’s IPUMS. 

The United States statistical system being extremely decentralized is useful to both the truth and to mistruths. On the one hand, putting all 115 data-generating institutions under the President’s aegis would be extremely hard and time consuming, and would require scalpel work on the bureaucracy rather than the sledgehammer Trump is taking to it. On the other, lower level administrative staff just lying to their superiors is a frequent source of bad data (China’s “lying mouths”, for example). 

The main problem for preventing Trump’s data manipulation, if he so desires it, is that the central node in the network of official statistics goes through the Office of Management and Budget, which is currently run by a camarilla of Musk lackeys and the explicitly fascist author of Project 2025, Russell Vought. And Musk’s indiscriminate staff purges are another massive issue: blowing up the statistical bureaucracy just blows up statistics themselves: the UK’s Office of National Statistics is still haunted by its 2015 decision to move agency operations to Wales (and other post-COVID issues like lower survey response rates) which led to swathes of experienced staff leaving the agency and reduced the reliability of their output. 

So in general, it’s really difficult to tell whether a Trump/Musk numbers-fudging operation could even theoretically be pulled off—especially given the, well, stupidity and incompetence of DOGE staff, particularly when faced with byzantine and arcane government systems.

Conclusion

In summary, it’s not that hard to tell that government statistics are off. People usually just come out and say it, or notice it—there’s unreasonable purges of the civil servants putting the indicators together, multiple indicators seem fundamentally at odds with similar ones, cross references, or just the very laws of economics, and people are not responding at all to the statistics (unlike the vibecession, where they just… responded inexplicably).So, all in all, the question is not necessarily whether Trump and Musk would want to deceive people with false statistics—it’s whether they would get away with it.

But the strategy of denying data to the public and experts is not just about muddying the field of discourse; two master liars like Musk and Trump know that the point is not to get people to believe fake numbers—it’s for them to question whether “real numbers” exist at all. Degrading the US’s statistical capacity, regardless of whether the numbers are better or believable, is an end in itself—lying, to fascists, is an exercise in power, not deceit.

Musk’s war on expertise and authority is simply an extension of his broader project to astroturf a mass reactionary mob—a sort of McCarthyism against the very idea of professional government. The reactionary turn against “universals” needs a joint assault on the very notion of a shared, objective reality—to replace a joint “real world” ruled by experts and their determinations with one of online mobs ruled by incoherent, disembodied “takes” and “vibes”. Without an official, believable CPI rate, inflation is whatever anyone says it is—and for half the United States, it will be what Trump says it is. 

Featured image is painting of a scene where Sigbrit reviews the custom accounts together with king Christian II of Denmark, by Kristian Zahrtmann