Lies and Statistics

Lies, damn lies and statistics.

I see many posts on Facebook claiming “evidence” for things that are simply not true. I am sure that someone has devised a statistic to enable the claim to be made but statistics can hide the truth as well as indicate one possible truth.

Statistic is defined as data capable of exact numerical representation, such as the correlation coefficient of two series or the standard deviation of a sample.

Statistics can be defined as a science concerned with the collection, classification, and interpretation of quantifiable data.

The last definition demands a “scientific” interpretation of “quantifiable” data. So much is published, that claims statistical truth but has too small a data base and is interpreted in a way no real genuine scientist would contemplate. True science requires understanding of all the possible variables and then building a hypothesis or (later) a theorem that includes these variables. Asking a hundred people in upper New York if they regret voting for President Trump is not science based statistical evidence.

Claiming a reduction in living standards due to British Conservative party governance, is equally a supposition, not an evidence based fact. To properly examine the effect of any government economic policy, you have to have a complete understanding of the economic state of the nation when they took office. And this has to by an accurate understanding, one not based on the propaganda of the previous administration. The base line you then start from, has to include all the stuff the previous administration had left in abeyance or put “off book.” To correctly judge any previous economic policy, you need data that can only be obtained some years after the ending of that policy.

Most of the misleading statements made in the media and on social media, use selective parts of a statistical analysis. For example a serious study on changes in living standards over a defined period of time; will include starting points, variations and evidential data for specific periods within the time period. It will not just include incomes for specified groups but also living costs. It will also state when events that affect the data take place. An obvious example is; if you were compiling data on changes to living standards of hourly paid workers between 1930 and 2000 then the second world war and all its ramifications would have to be included. In more modern terms any economic study of the period between 1980 and 2018 would have to include reference and allowance for, the growth of online and mobile technologies.

How good is the data used? For example if the study was on the number of people smoking in 2018. If you ask 1000 people; do they tell you the truth? If you study the sales of cigarettes, how do you account for the illegal sales of smuggled ones? If I take a sample of 5 acquaintances, one will say they smoke, while two more would say they do not—because they do not think the couple of cigarettes twice a day counts as smoking. One of my samples uses roll up tobacco that possibly has never see the inside of a customs shed. How good is your data?

How is the data presented? For example, if you take the number of deaths involving motorised vehicles between 1900 and 2000, you could say the statistics show a vast increase in deaths. If you study the number of deaths per vehicle mile, involving motorised vehicles between 1900 and 2000, you could say the statistics show a reduction in deaths. You can do the same for deaths in aircraft accidents, as long as you omit deaths during the various wars.

Political spin doctors, who used to be called press officers or more accurately propaganda merchants; are the real experts at selective statistical analysis. The press releases and private briefings they give selected journalists can be very beguiling but also very misleading. If the politicians start to believe the stuff given out by their own spin doctors, then this may have very serious consequences for a nation. Decisions can be made at political governance level, that are not based on honest, accurate, verifiable facts. People could end up paying for these sorts of mistakes for generations.

Consider the laws and strictures imposed regarding alcohol consumption. This is another mine field, it seems all the figures and “noise” are from people opposed to the consumption of alcohol. I saw claims that the figures show 2.78 percent of Britain's population have very high risk drinking levels; while Sweden only has 0.02 percent at high risk. Since there was no total investigation of every person, with each individual having the alcohol level in their blood tested, these precise figures are misleading. The United Kingdom is said to have a population of about 65 million, while Sweden is said to have about 10 million. So it is claimed close to two million UK citizens have high risk drink habits, while only 200,000 Swedes have the same risk. How were the figures obtained? If it was by questioning a random sample of people, how did they check the accuracy of the replies they were given? The general idea may well be correct but the precise figures are not verifiable by evidence based science. A larger concern is that the base level, the amount of alcohol consumed by an individual that is considered to be high risk, is an arbitrary figure. An assumed level that causes risk of harm. It makes no allowance for the age, size, diet, experience or metabolic rate of any person. It is another imposition of a one size fits all, propaganda figure.

Making laws or deciding on economic policy, based on this sort of data, is high risk.