Like do federal agencies such as the CDC, NASA, NOAA, etc publish fake data/statistics?
Given the amount of data they process and the number of people they employ I suppose at some point in their history someone probably did publish some fake data, but not generally, no.
Why wouldn’t they fabricate data?
Why would they? 
Er, ethics? Decency? Morality?
What is “the government”, anyway? If an agency of the executive branch of the United States federal government is fabricating data, and they’re doing so because they’ve been directed to do so from the very top–that is, by the President of the United States–well, we have a whole other branch of the federal government called Congress, which has oversight power over such matters. For much of President Obama’s administration, the Republicans controlled at least one house of the Congress, and they would have loved nothing more than to catch the Obama Administration doing such a thing. Now, we have the Democratic Party in control of the House of Representatives, and you can be sure they will be thoroughly investigating any possible shenanigans by the Trump Administration. Even in times when one party controls the White House and both houses of Congress, we’re still dealing with a bunch of ambitious politicians, who can’t necessarily be counted on to carry water for a President of their own party, since a good many of them may deep down be thinking “I should be in the Oval Office, not that guy”, would probably love the publicity of taking down a big scandalous executive branch conspiracy, and might even be genuinely interested in serving the public good and upholding the principles of representative democracy. (Granted, “divided government” does make it more likely Congress will zealously pursue its job of overseeing the executive branch.)
If an agency of the executive branch is fabricating data and it’s not being directed from the top, then in addition to Congress, they’re also lying to their boss (the President of the United States) who presumably will not be happy about that, and who ultimately commands all sorts of law enforcement and investigative agencies to ensure no such funny business is going on. In addition to the FBI and the Justice Department, all major divisions of the U.S. government (cabinet departments and major independent agencies) has an inspector general whose job it is to make sure everything is on the up and up. Even if the President is in on the fake-data conspiracy, that’s another set of eyes which must be closed and mouths which must be shut.
Also, the “executive branch of the United States federal government” is an enormous collection of people, many of whom are career civil servants who very likely are more loyal to the United States and its Constitution (or even just to their specific agency as a permanent institution) than they are to any given politician or set of politicians; all it would take would be one whistle-blower and everyone concerned would be looking at an FBI or I.G. investigation, along with hearings before Congress (conducted by the aforementioned ambitious and/or principled politicians).
And if for some reason the President, the inspectors general and other internal watchdogs within the executive branch, and both houses of Congress all fail to act, we still have a free press. One whistle-blower talking to the Washington Post and the New York Times could blow the lid off the whole thing (which in turn could essentially force such institutions as Congress and the Justice Department to do their jobs).
Fabricating statistics would be risky, and why would a civil servant do such a thing? In general, people who go into government service are honest and idealistic (at least initially). The ethically challenged and money- or power-hungry go into politics, which is where the money and power are.
But I have seen government reports that put a spin on the interpretation of statistics–shame on them.
Also, some government statistics are just plain wrong because of problems with the underlying data and human error. For example, for various good reasons there is gross under reporting of many diseases. Worse, the extent of under reporting can vary over time and from place to place. As for human error, I remember once when someone switched the headings on two columns of disease statistics and no one caught the error for a month or two.
One of the key reasons data is published so it can be used by others, often much later and for purposes that the original compilers may never have considered. That is, in itself a robust and very difficult to circumvent form of data verification, as its may be looking at long-term trends, pulling in multiple data series and using computer number-crunching.
If President Nixon ordered the Widget Measurement Authority to fake its counts of left handed widgets to stymie the commies, then any long-term graph would immediately highlight an anomalous period of widget production requiring explanation. A curious investigator would then go to other independent sources of government data to see what was going on, such as trade export figures for widgets, in case there was a big demand emerging there, and employment stats for widget factories, and perhaps even filed annual reports of the widget companies.
A conspiracist could then conclude that they are all in it together, but there is a point where that interwoven universe of facts and sources requires that Everyone is in on the conspiracy, except them. And there are other terms to describe that sort of perspective.
Those are scientific agencies, and what they publish is true to the best of their knowledge. Like all science, future research is likely to show that current models are incomplete, or in other language, “wrong.” But that’s how science works. If NOAA publishes a weather forecast and it doesn’t come true, that isn’t fake data, just data with limited accuracy. “Fake” to me suggests an intent for deception, and sure, there might be some instances of scientific misconduct, but the nature of science is that those types of things are often found out, because the experiment can’t be reproduced, or the analysis doesn’t stand up under review.
Intelligence agencies like the NSA and CIA are a completely different story, though. They have published reports saying, “we can’t do that” or “we didn’t do that,” only for later leaks or evidence to show that those were lies. I’m not talking about deception to cover up clandestine operations, but deliberately lying to those who are supposed to be overseeing their operations. They have been caught lying to Congress in reports and testimony many times.
Others have already explained fairly well, but I would say they would because, well, they’re the government…It’s not like they would get in trouble for doing such a thing.
Not American, but the Greeks famously did so with respect to their public finances,
How do you suspect publishing fabricated data would benefit the government?
No one here actually explained anything well for the fabrication of statistics.
The cited lying about the spying programs is something fundamentally different - these are of course clandestine programs
The various individuals and the component agencies can indeed get in the legal trouble in any democratic country and even those that are not democratic, when caught in violating an actual law by the enforcement authorities. And go to jail.
so the idea they would not get in trouble is a false one.
I am familiar with instances where the Cour de Comptes (the Audit Court in the Francophone, French law systems) has caught state enterprises and agencies in trying to fake numbers and the responsable officers - functionnaires have gone to jail and others at lower levels have lost their jobs and been permanently banned from the civil service.
This is indeed a good example of the difficulty of such.
The Greek finance ministyr and the statistical office worked together to fake certain kinds of statistics on the financial ratios for the Greek State. Not merely getting wrong, but actual active development of the fraud to allow first the ascension to the Euro and then later to fake the respect to certain funding ratios to keep the access to European funding and avoid EU commission penalties. This involved however not merely fabricating some numbers, but actually executing off-the-books swaps with some American bank: Goldman Sachs, with the clear intent of fraud.
Even with a high-level and major effort at faking these statistics, other analyses raised questions and finally the EU auditors uncovered it (or proved, it is perhaps a better word).
For the statistics that are heavily used by the parties other than those developing them, either other government agencies or the private parties, there is double checking, examination etc. As the most important of the government statistics are the time series statistics, and they are heavily examined by many parties, it is very difficult to consistently execute faked numbers. One maybe can get away with it once or twice - so a punctual study - but the time series data is heavily analyzed and the statistical tools and computer data processing power is directed to detect human manipulated numbers as there are many tell-tales that show up.
It is already the case for example for the PRC that it is well understood certain kinds of their statistics are fabricated - it is detectable and traceable. But of course it is the PRC - however in the investment banking there are the “shadow statistics” as even faked statistics in order to be coherent over time are usually in relationship in a structural way with real numbers.
It is in short virtually impossible to get away undetected with faking the time series statistics for any of the data sets that are published regulalarly and are used for their own interest by many other actors.
And of course the middle to higher level officers of the government bureaux that work on these numbers know very well that other agencies and the private sector - particulaly the big corporates and the big investment banks but also the specialized private data analytics companies are watching and have the mathematicians and the computer power to detect anamolies. So to fabricate is to run a huge risk in any of the bigger developed economies.
Even in the developing economies, there are a lot of actors watching and the oversight of the IMF detects “anamolies.”
It is very naive to think statistics can be just fabricated.
(this is a different question from can the statistics be wrong or have lazy errors, as the random lazy errors are actually much harder to detect than the “motivated” errors.)
“The government” doesn’t actually do… anything. It’s made up of people. And the people working for the government in scientific fields are scientists. Scientists who typically care about their reputation. And of course there are punishments for fabricating data. Your reputation is destroyed. Scientists who fabricate data are never taken seriously again - at least, within scientific circles. So why would any given scientist fabricate data for the government’s sake?
They certainly suppress and distort statistics they don’t like.
It’s unlikely in the vast majority of cases anybody straight up fabricating numbers, but there’s likely a lot of things like p-hacking without the appropriate corrections going on. This isn’t unique to the government though, and even takes place in respected institutions. Often due to the pressures regarding timetables for getting things out the door, or job security concerns for appearing to not be “doing” anything.
There’s also dodgy data collection mechanism and other ways to silently exclude groups (or artificially inflate numbers you want to look big). The definition of rape, for instance, is notoriously known to be sticky and not actually measure what it’s supposed to measure (the relevant federal definitions got better a few years ago but still doesn’t account for a few acts most people would term “rape”). Some of these aren’t intentional, of course, but I’d wager there are quite a few politically motivated definitions of various statistics you’d only notice if you looked at the fine print of some data and/or were an expert in some area.
But again, this isn’t unique to the government. A lot of people want their data to “tell a story” and often this sort of definitional data hacking happens in an entirely accidental or even well intentioned way.
E: Hell, a lot of gender breakdowns are subtly wrong because of the federal non-recognition of non-binary people, or third gender people (as well as the forcible assignment of intersex ppl at birth) for instance.
Necessary addendum:
No, this does not mean insert your favorite conspiracy theory is true.
This is not an example of ‘wrong’ in statistics but a policy level difference on the definitions where there may be a change of the consensus (or maybe not) on the definitions.
That is a different subject from the disortion or fabrication of numbers around the consensus (for the very subjective definitional) or the objective data.
Mixing the different subjects leads to confusion and unforutnately unfounded conspiracy thinking by mistaking differences of the interpretation or the definitional consensus for the fabrications.
Data may not be fabricated, but it is often worthless.
For example - the unemployment rate. It’s very common to have the current unemployment rate reported as x.y%, and then, 6 months later have the same agency report that due to additional data, or inaccurate sources or some excuse, the actual rate was a.b%
This happens enough to make me think that the statistics (especially ones that are “seasonally adjusted”, or otherwise massaged in some opaque fashion) are just pacifiers. The actual numbers are unreliable and unverifiable.
I argue that such corrections imply another reason. Collecting good data takes a lot of time. Unemployment figures are an excellent example. The data come from a public survey-corrected and validated by subsequent employment and unemployment reports. When the survey data is released there simply isn’t the data necessary to properly analyze the survey. They try, but as you point out they often get it wrong.
Such corrections are not necessarily a reflection of the failure of the method, but could be an indication that an unrealistic reporting schedule is imposed on a data collection effort. IOWs, the fault isn’t in the data it is in the reporting requirement.