Or, “return of the black helicopters”
Plenty of examples of hyperbole in current policy discussions, but here I want to return [0] to the specific topic of whether several key data series examined by economic analysts can be trusted, or whether in fact they are deliberately manipulated by government bureaucrats. Case in point is Econbrowser reader DickF‘s comments:
The government thinks it can run the economy on data that is years old and inaccurate at best. Also any time numbers are manipulated by government there is a political element involved. The whole reason the numbers are manipulated is to the will be “more normal” but who decides what is normal? In the government political bureaucrats who know their jobs depend on pleasing the politically connected. This is just another reason why centrally planned economies always fail. The hubris in government economic circles is enormous.
…
I am not saying that the agencies are manipulating data to make “each respective Administration look good.” Sometimes they manipulate date to make an Administration look worse than it actually is. It depends on their political inclination.
Here are some data revisions for GDP…
Figure 1: Growth rate of real GDP, SAAR, from 2008Q1 advance release (green), 2009Q1 final release (red), and 2009Q2 advance release+benchmark revision (blue). NBER defined peak at gray dashed line. Source: BEA, NIPA releases.
I’m cynical enough to believe that numbers quoted by governments and private companies are selected so as to put themselves in as good a light as possible (after all, the Bush White House was happy enough to sit on a global climate change report for a couple of years, and to defund statistical programs that measure the behavior of low income households). But this view is distinct from the perception that the statistical agencies that collect US national income and products account data (the BEA), the CPI and PPI and employment statistics (the BLS), industrial production and capacity utilization (the Federal Reserve), and import and export data (BEA/Census) purposefully and consciously distort and manipulate the data for their own nefarious purposes.
In my view, the reason why so many hold onto these views is because it’s so much easier to remain ignorant, and leap to the conspiracy view, than to do the hard work to understand why the statistics are imprecise measures of economic concepts, and why they are revised over time. After all, the former requires nothing more than taking somebody’s word, the latter entails reading the supporting documentation, comprehending what the terms used mean, and applying some basic math and statistics skills…And if that’s too hard, there are many people trying to help (For instance, Jim Hamilton cited some work by Oldprof on the BLS birth/death model in his last post).
As I noted in my previous post (The Government’s Macroeconomic Series: X-Files, Dilbert, or Resource Constraints?), there are (at least) three different categories of explanations for the shortcomings of government statistics. One is the conspiracy view (X-Files), one is the incompetence view (Dilbert — although upon reflection, this should be called “pointy-haired boss”), and one is a realization that government statistics are generated by organizations with limited resources, facing legal constraints on data collection, and confronted with an ever-changing structure of the economy.
So, if one wants improved government economic statistics, then tell your elected representative to keep up support for proper funding of the statistical agencies (time series here) FoxBusiness has some reportage.
See also Mark Thoma. Dean Croushore has a good review of why macro statistics exhibit the characteristics they do over different vintages.
Paranoia is the easy way out for those who refuse to question their own beliefs.
There are enough people involved in gathering and analyzing the data that it would be difficult to fudge the data.
I wonder if the same can be said for China and Russia?
However, there is an interesting subplot – – – I was watchig CNBC on Friday before the jobs report and they wondered if the better than expected employment numbers had been leaked early to the White House based on comments by President O’Bama the day before. At the same time, they noted that the night before the release, Goldman Sachs reduced their forecasted change in employment to 250k.
I have often wondered if the BLS and others give the FED and White House an advance peek at the data before a big release. I don’t have a problem with that per se, but I can see where some of the FED or White House insiders might then leak the numbers to the likes of a Goldman who can push the inside info out to select clients.
I think many people deliberately feed the paranoia with counter knowledge for their own financial gain. Big newsletter mills such Agora Financial and Investorplace just laugh all the way to the bank.
Let’s put the discussion in the right context.
Expectations of growth and inflation can adversely affect the real economy. Those adverse impacts affect the general welfare. The government is charged with maintaining the general welfare. Ergo, the government must be concerned with creating favorable expectations of growth and inflation.
What are the means available for managing these expectations? There are the traditional levers — fiscal and monetary policy. After that it gets a little murky. Should the government ban short selling? The answer has been “yes”. How about, as a policy maker, giving consistently rosy outlooks? Presumably this is commonly done. Using the Fed’s balance sheet to manipulate market prices and generate price signals? Its being done in spades in the credit markets. How about a Fed Board that consistently downplays the long run costs of monetary policy? Is this allowed?
And then there’s the possibility that political appointees shade the data results presented by bureaucrats — data which is inherently open to interpretation and judgment calls, such that the shading doesn’t necessarily rise to the level of Argentine-style fraud. Menzie thinks this is close to impossible, and yet there is evidence from the Nixon administration that this was done.
My point is that there is a spectrum of lying that can occur in the interests of maintaining the general welfare. Where will the government choose to stop on that spectrum? I would not be complacent about the answer to that question.
David Pearson: There’s a distinction between looking at GDP numbers as reported, and then interpreting them in as favorable light (if perhaps injudiciously so), and actually telling the agency in question to depart from standard procedure and up the reported GDP figures. The former is done on a regular basis. I’m waiting for an example of the latter in the macroeconomics sphere.
By the way, I don’t deny that this sort of distortion might happen in other arenas. See for instance the Bush Administration treatment of the head of the Bureau of Justice Statistics in 2005.
I am not an economist and not until the crash did I realize how truly mystical math can be.
Some of the very best and brightest minds in the company used “fail-proof” math models to prove that bundling up bad mortgage loans into large bundles of securities to sell to pension funds and the like would be a highly profitable venture.
And it was highly profitable, until recently, when the loans were recognized for what they always were – toxic.
And we just this week heard the loud cheers of celebration because the unemployment rate dipped to 9.4 percent from 9.5 percent – but if you looked at the numbers, it’s because people stopped looking for work.
Economists can make numbers do anything they want them to do. Math today is all magic and misty – but not very trustworthy any more. Most people outside of Washington don’t put much faith in the figures policy makers toss out.
Political influence on economic statistics in the US occurs frequently in the form of policy changes. Manzie’s mention of defunding potentially politically damaging data series is one example. The most relevant for economists are methodology “improvements” that are made from time to time, which tend to increase politically positive numbers like real GDP growth and decrease politically negative numbers like inflation and unemployment.
I doubt there is much motive for the White House to demand that some official economic statistic publisher wiggle one month’s figure, especially given the risk that the manipulation would be exposed.
PS I don’t see any particular rise in the trend of grand-conspiracy theorizing. It seems to me to be something that is naturally built-in to the human psyche, a natural defense againt the highly conspiratorial nature of human beings. Like chimpanzees building alliances and waiting for opportunities to overturn the current alpha, every human hierarchy is full of mutual suspicions and intrigue. Of course, there’s a big difference between small conspiracies, which are commonplace, and grand conspiracies, which are imagined by nutters. But it seems to me that people are just naturally suspicious of foul play, and they will often act on their suspicions without thinking them through enough to realize that only a grand conspiracy could accomplish what they are suspecting.
Paranoia is I think a different, biological condition, which weakens rationality and heightens suspiciousness.
@tj :
“I have often wondered if the BLS and others give the FED and White House an advance peek at the data before a big release. I don’t have a problem with that per se, but I can see where some of the FED or White House insiders might then leak the numbers to the likes of a Goldman who can push the inside info out to select clients.”
your assessment is correct. selected clients of gs get insider info, and that is not paranoid or based on conspiracy theory, but reality.
information rules, and that hank paulson called a few of his buddies at gs – who cares ?
While I agree that BEA, BLS, & the Fed are not purposefully and consciously distorting and manipulating data for their own nefarious purposes, I submit that they may well be doing it for their political masters.
I have tried to track the changes in methodology used to report various measures, among them CPI (“hedonics”), U3/U6–no longer looking, part-time, GDP, M3(discontinued). John Williams at Shadow Government Statistics (www.shadowstats.com) does a far more thorough job than I have or could.
My conclusion from this informal process is that, over at least the last quarter century, EVERY change in in USG economic measurement methodology has made the then-current economic situation look better than the previous measure. (It may well make earlier periods look less successful, but that’s obviously not the POLITICAL point.)
As a result, I would suggest that’s is not paranoid to think that the USG is manipulating economic data, especially in the face of evidence that it has. Of course, each change is presented with a smiling economist face about “improving” measurement. Well, maybe not so much.
Moreover, I put this challenge to you: Provide evidence that ANY official US economic statistical methodology that has been changed in the last quarter century HAS NOT made the then-current economic situation look better. You haven’t don’t so in this article and, indeed, resort to namecalling–“ignorant”, “conspiracy view”.
Show me the money.
This is a pretty good summary, though I think you are a bit too trusting.
Sometimes data seems inaccurate because of arguments over method. For instance, I think that the CPI is too narrow and thus doesn’t measure inflation properly. However, that’s just an opinion. One could say the government is always predisposed towards picking from a range of possible methods the one that places it in the best light, but that’s not exactly a smoking gun of fraud.
Unfortunately, some work the government does seems like little more then out and out fraud to me. The best example I can think of is the BLS birth/death model. A model that consistently adds far to many jobs, is constantly revised down by large amounts, and gives nonsensical results like increasing employment in construction and business/finance during this downturn. Simply Dilbert syndrome is insufficient here, as the model has been so wrong for so long that its continued use is an implied endorsement of its method and results.
Terry: Please see previous discussions on Econbrowser regarding John Williams’ Shadowstats: Shadowstats debunked Shadowstats responds.
If one’s assertions are the statistics are manipulated to make the current administration look good, well look at Figure 1 and see if that jibes with the thesis. If you are saying that methodological improvements are likely to occur when convenient, then you aren’t too far from Mark Thoma’s view linked to at the end of the post. (I do wonder if every change fits the model; chain weighting GDP components was planned long in advance of implementation.)
dave: On the CPI issue, see [1], [2] and the links contained therein. On the birth/death model, see links to Oldprof in Jim Hamilton’s post two days ago.
The entire debate has been diverted from the most important point that should be made. As Menzie admits statistics are often and at times significantly “revised.” I thank Menzie for quoting my sentence that is most important.
“The government thinks it can run the economy on data that is years old and inaccurate at best.” And this makes no difference whether it is intentional or simply the nature of such statistics.
Our current economic problems are because government officials and politicians have the hubris to believe they can run economies better than the people. They believe that they know better what is good for us than we know ourselves. They believe that they can gather statistics that will give them more knowledge and allow them to make better decisions than the millions of transactors in the marketplace who cooperate to make something as simple as a pencil http://www.econlib.org/library/Essays/rdPncl1.html
If you want to debate human nature in fudging numbers that is fine. Such fudging will go on until the end of the human race. That is the nature of man. But for me the more important issue is to prevent, as much as possible, number-fudging busy-bodies running my life. Enough.
There is an unfortunate tone of smugness to this post.
My prior is that people who read econbrowser are a pretty geeky econ crowd (it’s a 5 calculator for good reason). I’m therefore really surprised that the implied selection mechanism allows anyone with this degree of paranoia to filter into the comments. Chain weighting? Seriously?
On the other hand, maybe paranoid people are more likely to leave comments. I have no idea. I really should have paid more attention in Micro. God I hated that class.
Menzie,
Thanks for this thread BTW.
Thought you might find this interesting.
http://whitehouse.blogs.foxnews.com/2009/07/22/wh-pushes-back-against-cbo-meeting-criticism/
Doc, a data series — CPI — that does not capture asset inflation — the home price bubble — and is used far and wide by The Fed and federal government over the years to claim there is no inflation prompts incredulity by perceptive people.
You call such paranoia. I call it reasonable disbelief.
And, why you grant great deference to a privately-owned entity — the U.S. central bank, a.k.a. The Federal Reserve Bank — is dumbfounding to me.
Bernanke may not be a crook, himself. But, the outfit that he runs is owned by crooks.
A good example of the paranoia that Menzie describes is featured near the end of the following NY POST column (in reference to the BLS’s birth/death model):
http://www.nypost.com/seven/08112009/business/obama_does_a_good_job_tipping_employment_183967.htm?page=2
It would appear to be an easier task to believe what that guy wrote than to read and understand, e.g., the discussion that Jim Hamilton linked to the other day from OldProf.
There are lots of statistics that are fake and that are used for political purposes. Like the “fact” that there are 50 million uninsured people in the country.
Is the CPI or unemployment rate or # of jobs created rigged? Probably not. And even if they were, and even if the Obama-loving media hyped them to high heavan (which they would), would that really change our perceptions of what is going on out there?
We know the people that we know. Are the unemployed, or have they recently found work?
We live where we live. Are houses on our streets selling? Are they being forclosed on?
We shop where we shop. Are stores going out of business? Is there a lot of vacancies?
Really, the conspiracy aspect of this is a whole lot of nothing.
On the other hand, DickF’s main point, that you can’t expect to run an economy on such bad data, is totally true. Less is more on the fiscal and monetary fronts. You can’t fine tune a modern economy.
Menzie–Responding to your reply:
I have read your critiques and others of Williams’ Shadowstats (including the BLS’), and I’m not saying they’re perfect. But he sure has been at this a lot more methodically than me.
Re Figure 1–I’ll defer here to Jake at EconomPicsData who noted that had 2Q09 GDP been calculated the same way as 1Q, the annualized rate of decline would have been 5.8%, not the one percent generated by the new method. Moreover, Jake’s GDP number is absolutely consistent with the 6% annualized decline over the previous five quarters.
I guess I’m just more cynical/realistic than you. For the record, I’ve done analysis for US policymakers, and seen how they have tried to influence the direction of that work a priori, and then how they perversely used the results. I’ve earned the right to be cynical, or maybe it’s realistic.
BTW–the challenge still stands!
Menzie, I think you are making a little bit of mistake here of being overly judgemental of DickF and comments of similar nature. I think that usually such comments are the result of an inablility by people to articulate much more subastantial, but complicated, reasons why measures of subject are poor measures in general, not really for the reasons stated. Also I think they result mainly from frustration with people puting to much value on the measures that have much little meaning, especially in the short run and for people outside the finance industry, in the current environment.
For GDP in particular I think that interest expense and rent probably should not be counted. It’s really just a wealth transer.
Menzie,
My problem with the CPI is not of the shadowstats variety (I find his work a little shotty). Its more a matter of believing that inflation manifests itself in things other then consumer goods. Namely, I would consider something like the use of debt to big up house prices as an inflationary event. The Fed doesn’t target asset prices, and the huge increase in debt (which is basically money in our system) during the housing bubble didn’t show up in CPI. Now that the bubble is burst you have a huge deflationary effect, but once again it is not reflected in the CPI (except inadvertently).
So the effect of using CPI of consumer goods only and not considering debt and assets means that real interest rates were far below “inflation” 2002-2005 and are now extremely high, as “inflation” or more accurately deflation, and far below zero currently even though the CPI is around flat.
The actions of treasuries and financial markets make a lot more sense when you consider inflation more generally rather then zooming in on consumer prices, which is part of the error the Fed made.
For Birth/Death issue I have a qualm.
First, the linked author examines the birth/death adjustments on a yearly basis. However, within the birth death yearly change there are several monthly changes. Typically this is about 10 months of very rosy additions and two months of negative benchmarking revisions. Those monthly additions end up making it into the big headline number that makes the front page of the NYTimes, but the revision months adjustments aren’t included big headline numbers (I believe, correct me if I’m wrong). There is good reason not to do so (there are six months of revisions in there), but that means the other 10 months headlines are getting boosted. Perhaps the revised numbers do lead to a more accurate jobs picture on a yearly basis, but your still talking about PR management if the headline numbers on newspaper stands are being manipulated upward.
Terry: With all due respect, I don’t understand Jake’s calculation. I don’t see how it makes sense to compare a pre-revision number to an advance number; shouldn’t we use the most comprehensive estimates in our calculations of q/q changes?
I’ve also worked for policymakers. I think I have my share of cynicism/realism, but I think I have my cynicism directed at a different place than you do.
Most economic statistics are about as good as one can expect. A very large industry of analysts scrubs and compares every number, so its really hard to see how purposeful distortions could last for more than a month or two, even if they were created. With so many other sources of uncertainty – who cares?
The CPI treatment of housing and the employment Birth/Death model are the two major problems because of methodological deficiencies. If you lived in NYC, Miami or LA and tried to buy housing in the past few years, you know that the CPI number is wacko. Many other areas of the country are less extreme examples, and people in those areas “have their suspicions” about the housing part of the CPI. Now that house prices are down 20% or more in most of these areas, you would never suspect that from the CPI housing numbers. Common experience tells millions of people that something is broken in CPI housing, so what else is broken in CPI becomes the next question.
The Birth/Death model is more complex. I read the 3 articles that JDH linked, and though I consider myself well educated, I found all 3 of them hard to understand, unpersuasive, and poorly written. Though I learned that careful statistical techniques are used, when I step back from the trees and look at the forest, I see a trend following algorithm. Trend followers are always wrong at the turn points. Its just math. They do not purposefully distort, they just have a methodological deficiency in turning point times like the ones we have just been though.
I had offered my criticism earlier that there were two options – inadequate metric or fudging of data. I’m now inclined to inadequate metric but since the labor force number is at least partially generated from surveys, I’ll remain skeptical. After all, polling is notoriously suspectible to sampling and question skewing.
I find your example of the Bush Administration siting on the global climate change report telling. It was a bureaucratic hit piece that should have been prevented from having a government seal attached. That was a responsible action.
Of course, the British government has avoided the problem by making temperature records for the UK classified state secrets.
Perhaps the skeptism that the unemployment rate number was greeted with is really about the cheerleading from MSM and academia. The slight rate improvement was clearly NOT reflecting the real hurt on the ground when labor market participation fell even more.
And yes, the Obama Administration has given me ample cause to be distrustful of their words and deeds. I will continue to mistrust them. Every citizen should.
Joseph Somsel: Er, IP, cap utilization, nonfarm payroll employment are also generated by surveys…
Mike Laird: There are many trend-following algorithms used in generating first estimates of many series. See here for a discussion of GDP.
I too work for elected policy makers. And, yes, they often want to skew numbers in a favorable light. But they are also much too busy to get into the details of how numbers are collected and compiled, so would not have the patience to develop and implement a scheme to systematically distort statistics. It is much easier for them to suppress unfavorable information or delay its release, without changing the way statistics are gathered and reported. So the “X Files” theory of data manipulation is not very realistic.
I also know a lot of folks like me, who make their living presenting quantitative information to elected officials (data nerd bureaucrats, if you will). Almost all of them have some kind of civil service protection, which makes them both ineligible and uninterested in political advancement, so pleasing “political masters” is for the most part irrelevant to their careers. Instead, they work hard to provide an honest assessment of the data they analyze and report, without regard to politics. This is not their devotion to the public good (although that can be a motivation) but pure self interest. A civil serveant who becomes too cozy with the current elected leaders can find their careers in trouble when the opposition takes over. The best move is to stay as objective as possible, so you are trusted regardless of who is in power. This reality is another reason to doubt the “X Files” theory.
Purely superficial comment here.
I think you should call it “Best Effort” rather than “Resource Constraints”. “Resource Constraints” explains the how, but it doesn’t really capture motivations in the same way that “X-files” or “Dilbert”/”Pointy-haired boss” do.
And I think various assumptions about government motivations is the essential dimension on which these options vary.
government statisticians are NOT manipulating data to make any government look bad. they will be charged with conspiracy against the state if they attempted.
by being serfs of elected officials, their role is to please and they bend to pressure and accomodate as much as they are asked to.
For Mark Freeman–I generally share your views on the experts/bureaucrats who put together statistics and analysis for policymakers. They certainly try to be meticulous and honest. The problem comes when their “technical” work runs against the grain of an important (not all) political position in an administration. While they do not usually risk termination, incessant refusal to adapt (almost Darwinian) to policy views virtually assures a negative career prognosis. No chance for advancement (at least during the then-current administration), shifted to an irrelevant position, etc. In the end, each of these experts’ work is overseen by a political person at some level–who will defend an administration’s view way passed the point of reason.
Terry–Well put. And a smart bureaucrat will not incessantly refuse to adapt to political realities. But in almost all cases, such adaptation does not require a deliberate distortion of statistics. The political leaders are not often concerned about what statistics are being gathered, or the methods used to gather them. They, instead, focus their attention on how the data is presented to the public (in press releases and other high-level reports). This also means that politicians do not pay attention to the detailed numbers behind the high-level numbers reported. So a bureaucrat does not generally have to choose between valid statistics and political realities, as long as they are willing to accept how the information is presented to the public (i.e., as long as they do not involve themeselves in crafting the spin). That said, bureaucrats often (and sometimes successfully) argue against a prevailing political view, using the data they have gathered. And I think you might be suprised at how often politicians appreciate the bureaucratic pushback, and change their views in response to the information presented. Bureaucrats are no different than anyone else in our democratic government: they too have to persuade and be willing to accept decisions they may not personally support.
As a former government bureaucrat, my impression is that numbers are distorted in a way that smooths out the edges. They don’t want to report every outlier as something real. So if the data look a little unexpected, the urge is to dampen the surprise. They don’t want to answer a bunch of questions.
The incentives in government are to be as absolutely boring as possible.
Politics has just an iddy bit of influence. But not much.
Think extreme conservativism. As the word really means, not its political connotations.