There are several models for judging and documenting the accuracy and Quality of official Statistics. 

https://www.google.com/url?sa=t&source=web&rct=j&url=http://www.ssb.no/omssb/tjenester-og-verktoy/kurs-og-seminarer/nordisk-statistikermote-vedlegg/_attachment/130649%3F_ts%3D1402f6cef38&ved=0ahUKEwjKkPjFxa7KAhVSA44KHSLMAtM4HhAWCB8wAQ&usg=AFQjCNEFNFCuZsArn2aRP-7r5eIunsCNYw&sig2=0yiUuxplFGuD6jQiTCpVJg

official accuracy of course in Indian context

Subrata, do you mean that official statistics should be held to a high standard?  My impression is that statistics are held in high regard in India.  Or do you mean that, as everywhere, politics can be in the way (causing bias)?  Further, in perhaps most organizations, internal agency issues can impact efficiency negatively (causing increased variance, and maybe a bias).  I'm not clear on your meaning here.  There may be intentional inaccuracy, but I had not thought of that. I was thinking about unintentional inaccuracy, though I imagine there may be both in many official statistics, worldwide, and that is a concern. 

There are several models for judging and documenting the accuracy and Quality of official Statistics. 

https://www.google.com/url?sa=t&source=web&rct=j&url=http://www.ssb.no/omssb/tjenester-og-verktoy/kurs-og-seminarer/nordisk-statistikermote-vedlegg/_attachment/130649%3F_ts%3D1402f6cef38&ved=0ahUKEwjKkPjFxa7KAhVSA44KHSLMAtM4HhAWCB8wAQ&usg=AFQjCNEFNFCuZsArn2aRP-7r5eIunsCNYw&sig2=0yiUuxplFGuD6jQiTCpVJg

Dear James,

Statistics is one of the weapons people use to deceive society. As a responsible citizen, we have to demand data always be gathered fairly, processed on the right methodology and presented unbiased on a timely manner in order society to make right decisions.  A right decision is the one that is made based on the right information  on a right time. 

Dear all

Only the government with Credibility can hold a credible official statistic.

Dear James,

I tended for a while to like Chun's response, but I could not convince myself of my experience statistics presented by different government agencies at national and international level. Even credible governments set goals so high that to fulfill them is impossible. Therefore, to create a perception of progress and that goals set are achieved, these establishment too play with statistics. Even regimented establishments with efficient decision making and implementation cannot deflect this blame of tampering with statistics to fool the public.

James: your suspicion is in the right path. Because accuracy of Official statistics is subject to many levels from field data collector to the highest office. Firoz have some genuine points. Yes you are correct intentional  and unintentional accuracy both exist in the sysytem

Dear James,

Regarding Hungarian context, university teaching of statistics is weak. Most students have no idea even on elementary statistical procedures or opportunities. Many average scientists think statistics is only a kind of decoration.  Good statistical books indicate clearly the limits of statistical approach. My only wish is to have fine statistical teaching at universities and to show how to use efficiently statistics. Statistical application has two weak points: some think this is only a “decoration” of data and some mystify statistical results and cannot acknowledge what Georges Box put it that statistics can provide only useful approximations and cannot be a dogma.  

Some will define accuracy as the degree to which something is true or exact. Knowing that there is a widespread corruption in many governments, official statistics can be played with to serve certain political purposes & biases. In a 3rd world country, I remember an incumbent president (then) taking the people to the ballot boxes to answer a single question: Do you support the president ? and Yes or No has to be encircled. The minister of the interior (then) announced, after the "democratic!" poll, that those who gave "Yes" to the president were 100%. This big lie was a big joke in that country as was reported to the president so in the next ballot, after few years, the president was given 99.99 %. Increasing the number of digits has never indicated accuracy in that country.

Scientifically speaking, one has to doubt any statistics coming from certain countries. In some countries, they changed " true" statistics to "false" ones in order to say that the population of a certain sect is higher or to say that a certain politician is more popular or to say that the number of unemployed is less than the reality.

Thank you all for your responses. 

The following remark by Andras struck me in particular, based on my experience:  

"Statistical application has two weak points: some think this is only a 'decoration' of data and some mystify statistical results...."

I always thought it very annoying when someone at the govt. statistical agency where I worked would refer to statistics as "magic."  (This is not in the respectful way that Andras may have meant, but more dismissive, as in thinking of it as "decoration.")  Why would someone work at a statistical agency when they use the word "magic" to describe it?   And I found it annoying that though I tried to do as much as I could with as little as possible, that even a much more supportive supervisor thought that meant I could then do something with nothing, and was angry on an occasion when I said I could not. Another time I heard him announce in a presentation that we were basically going to report estimates for something, without collecting data. I asked him if I had heard him correctly, and he said yes, "You'll think of something."  (Luckily I talked him into collecting a little data.)  - A very massive increase in the commitment to provide information was made in another case, without management understanding the consequences, which I had to meet with a huge effort, after I found out about it.  Management told me, when I heard about it and asked, that they had "already decided."  (That was another one of the 'more supportive' managers.  It was far worse at much later occasions.)   They apparently thought it was all automatic.  So I was told it just had to happen.  That meant a huge change in methodology that I had to develop almost immediately.  (I mean months rather than years.)  

My problem is when people who should know better because of the positions they occupy, do not understand, respect, or even consider statistics to be important.  When you confuse being an industry expert, with being a provider of statistics on that industry, you have a problem.  You should know your subject matter, but you cannot provide statistics on it so well if you consider statistics not to be important, and do not understand them.  You cannot provide good management at a statistical agency, if you do not even understand the goals of a statistical agency, and the reasons for having such an agency.

Dear James et all, 

Unfortunately the reality of your company is part of realities of certain universities run by ignorant and full of power liders  The problem is that many who are close to that kind of leaders want to keep their jobs, so obey blindly and become accomplices of that going backwards in science. These "headless"  infect colleagues/students. Thinking is seen as a wast of time .If people had compassion, none of this would happen because the power of thinking for the good of all, would in the only admissivel and not the hierachical position ...

Statistics are only as good as the input, and their interpretation is only useful if they are approached with an unbiased mind.

For a decade Canada was in the grip of a party of ideologically steeped, intellectually withered clique who were close-minded enough to give legitimately conservative people a bad name. Their leader - unfortunately prime minister for more years than one wants to recall - was closed, inflexible, stuck in a distorted view of the 1950s and thoroughly drenched in the apocalyptic vision that posits an imminent war between "good" (evangelicals) and "evil" (anyone else, but particularly anyone against Israel as the personal fief of God, and Muslims of all stripes).

They denied global warming and much more, disbelieved in scientific approaches to solvable problems etc and committed a series of intellectual crimes: doing away with the long-form census and destroying scientific data and research materials and closing libraries (in one agricultural library staff rescued books from the dumpster even though it had been made clear that such would constitute a crime).

They made us a laughing stock by throttling all inquiries from the public and media.

When the new government came to power in early November t found shocking gaps in information. Census data are required for the orderly administration of any number of government-provided services - from schools to transportation to health, one doesn't have to remind you of the consequence of so egregious a loss.

Yes, one can "prove" that black is white using statistics, but statistics are also necessary and as such one must expect - and provide - the most up-to-date and sound statistics possible.

I agree with Dr. Helena Pestana. 

What information regarding accuracy would you like to see?  (I think that at the least, estimates for relative standard errors should be included.)  Metadata? 

What means may be found to hold the published data accountable? 

Starting form last: number of digits says nothing to me, after all nobody cares or knows about floating point arithmetics and precision in social sciences (these are small letters for them, I can ensure you about it from my epxerience: if you tell them about epsilon of the machine, they will look at treat you like an allien...)

What I'd like to see published are some variables that can measure the quality of interview:

• Did the participant answer the questions with interest?

Then I can estimate the overall quality of research.

Official statistics are manipulated and decorative. These need discounting and need filtering to reduce the biasness.

official statistics is fulfilling the requirement and biased as well as positive to show the organization is moving in right direction.

in my view, official statistics are just like indicators, it is highly impossible to have 99% accuracy and most of the figures are unweighted. I believe 60-80% accuracy we can expect from official statistics (variable wise). 

"A statistical analysis, properly conducted, is a delicate dissection of uncertainties, a surgery of suppositions."

M.J. Moroney

To give an unbias report at any point in time 

The statistics are only as valid as the data input. If workers are forced to find workarounds to enter data, or difficulty of data entry is high, the output stats will be affected. That is why it is critical to understand the workflows of the end-user when a data collection application is planned and designed.

http://www.pmi.org/~/media/PDF/Business-Solutions/Complexity_Dexterity_whitepaper_v2.ashx

Problems of accuracy in official statistics on drug-related deaths.

Shai D1.

Abstract

The accuracy of statistics on drug-related deaths depends on numerous processes at work from the investigation of the death-scene to the categorization of the case in official statistics. This article discusses problems in lack of information, pressures on medical examiners, definition, nomenclature, and conventions of coding. The implications for comparisons of rates of drug-related deaths across ethnic groups and geographic areas are discussed.

http://www.ncbi.nlm.nih.gov/pubmed/7890443

I think it depends on the purpose of the statistics and the accuracy of their data. If the purpose is a scientific study, statistics will be revised and approved or not by peers. If the purpose is political and made by official or governmental agencies, there could be some bias in the input data or their estimation for governments popularity interests. If statistics are conducted by independent agencies, there could be a % of true estimation and a % of weak one based on the statistics' methodology and approach as well as the degree of accuracy of the samples data used

In canda  ....statistics inputs ar regulate by the government 

What accuracy do you expect from Official Statistics?

Generally, all statistics involve some degree of inaccuracy, even if they are prepared and produced with maximum care.

Ideally, official Statistics must be accurate, unbiased, impartial and transparent. In reality, official statistics are biased and sometimes even manipulated to provide a better image of society/country in different subject areas: economic, demographic, social.

The approach adopted by the OECD Electronic Working Group to measure trust in official statistics is based on three factors:

Structural Factors – the extent statistics are (or are perceived as being) independent, objective, impartial (i.e. no political interference) and transparent (e.g. an adhered to advance release calendar, clear reasons for revisions, etc.).

Statistical Factors – complying with international standards and having sound  methodological practices, robust statistical processes and quality outputs.

Reputation – the factors determining the public’s opinion of official statistics can be influenced by a number practices undertaken by the compiler; examples include:

publication of relevant data that informs all users of issues of national importance; public consultation, including the media; handling of incidents or issues involving wrong data or unethical practices; etc

http://www.oecd.org/std/Stat%20newsletter%20final%20for%20the%20web.pdf

Don’t question my data, get your analysis right, says India’s Chief Statistician

http://www.thehindu.com/business/Economy/dont-question-my-data-get-your-analysis-right-says-indias-chief-statistician/article8136586.ece

Dear James,

Your question is somewhat intresting. In my opinion, for official statistics periodically produced from a government agency, it is better to be indicative of the accuracy or precision. In general speaking, if the accuracy of statistics is not provided, it should be more than 80%, even 90%. If the statistics are more inaccurate than the real changes, that is, the standard errors are more than the changes, then it imply the periodical interval is not long enough.

Good luck!

See ...

Nice figure, Subhash.  Where did you find this?  

Krishnan, above, has shown a good example of "Official Statistics."   It also points out that sometimes apparent problems may be with the analyses - while sometimes it could be the data - maybe both. 

I don't think there is a single answer.  There is data for which timeliness is a priority (a lot of economic data, for example), and the goal there should be to have the number of decimal places reflect the expected accuracy of the data.  Poll data is usually reported including margin of error.  Bias is often well-understood for things like unemployment statistics, but less clear in many other applications - especially polls, which are so sensitive to how the questions are worded.  In general, my expectations for (and trust in) official statistics are low, due to the many factors encompassed in Subhash's graphic above.