So, it is hardly controversial to say that people use statistics “creatively” to support their preferred argument. That is to say, people misuse statistics, because statistics as a tool has no such devious agenda. I think basic statistical training is rather essential to cut through the deceptive tactics.
Take for example, this argument from an alcohol industry group:
A tax on beverage alcohol is a tax on the hospitality industry – threatening the 3.7 million employees of manufacturers, wholesalers and retailers. Federal, state and local governments lose when excise taxes on spirits are raised. The impact is felt throughout the U.S. economy. The last two federal excise tax increases on spirits resulted in:
+The 1991 increase resulted in a reduction in total federal alcohol excise tax revenues for the succeeding 5 years.
+The elimination of 98,000 jobs generated directly and indirectly by the distilled spirits industry.
+Nearly $1.3 billion in lost wages.
+A total $1.1 billion reduction in personal and corporate income, sales, and property tax revenue to federal, state and local governments.
+States paid out more than $150 million in unemployment benefits.
An alcohol tax increase passed in 1990, and these things happened to alcohol related industry in the following five years. Thus the alcohol tax increased caused these things, right? This is the logical fallacy “post hoc ergo propter hoc” after which, therefore because of. We’ve got correlation and even causal logic, because certainly a tax on alcohol would hurt the alcohol industry.
So why is this problematic? Well, as it happens, the country slid into a recession in 1991-1992 and took a few years to recover. Unless they are arguing that their industry was impervious to the recession (which certainly was claimed previously) much of their pain was the result of the recession, not the tax.
However, aside from the statistics, there is a value question that can be addressed. When Congress raised the alcohol tax, they knew full well that some harm would fall on the industry (you get less of what you tax, and if you have less alcohol consumption the industry has less revenue and thus jobs). But the appropriate measure is not whether or not the industry suffers, but comparing that cost compared the the reduced social cost of alcohol use through lessened consumption. Congress did not raise the tax rate to benefit the industry or simply to raise revenue, but in order to gain this broader benefit. I do not have statistics on hand to show that the tax increase provided such net social benefit, but it is certainly suggested.
Statistics cannot address the question of values. The statistician’s goal is to take the premise and to establish whether it is accurate, and let the policy maker decide which values should prevail. If that standard were held, the tone of politics would be much improved. The problem is that too often, values and the statistical claims used to support them are in contradiction. Thus instead of claiming that we should lower taxes because taxation is wrong (value claim), we hear claims that we should lower taxes because it will do x, y and z (economic growth, increased revenue, etc) things that we value. This latter proposal is popular, who is against economic growth and increased revenue (especially when it can be done by cutting taxes), but using the faulty statistical claim allows this support to be placed in policies and ultimately outcomes that are quite different.
These claims reveal a political negligence at best and political fraud at worst, but there is no system in place to adjudicate such claims. Thus we are left with a postmodern world where no objective reality exists and everyone’s claim is equally valid. And if all of them are equally valid, one might as well favor the policies that sound easiest. Eventually we become a timid and naive public that does not tolerate the big changes needed to actually solve our problems.
