Benjamin Disraeli’s Revenge

Benjamin Disraeli was a fascinating man for a variety of reasons. As one of those impossibly accomplished 19th century figures, who was a successful writer of novels, as well as an immensely quotable and successful political leaders, who served multiple times as Prime Minister of England, he is a fascinating character for a variety of reasons. As an “ethnic” Jew who was a staunch Conservative, he steadfastly opposed Gladstone (his archrival) over and over again over decades of political history during the second half of the 19th century. Gladstone made most of his liberal appeals based on statistics, and it was Disraeli’s stinging retort that there were three types of lies: lies, damned lies, and statistics. Today I would like to talk a little bit about the third type of lies.

I am fascinated by probability and statistics. I’m not exactly sure where this fascination began, but I kept track of music chart statistics, and compiled my own charts based on a variety of other charts as a teenager, and so my interest must have started at least then, if not before. Most of my own statistical collections have been rather straightforward, but that is because most of my statistics have been made for myself without an interest in influencing others. For example, when I compiled the statistics for my music chart, I simply inverted the chart position of several charts–using the top 50 of the Billboard Hot 100 and a couple of other charts, along with a couple of genre charts on Billboard (like country, R&B, and rock) to create a comprehensive listing that took airplay and sales into consideration. When I compiled statistics for a picking group for NFL games, it was a straightforward collection of picks and collecting stats on whether they were successful or not and aggregating these throughout the season.

Most of the statistics that I look at are far more complicated than those I have compiled in most of my life. To be sure, when I was a student I had to do some coursework in design of experiments. I have also had to compile statistics on surveys, experiments, and naturalistic observations of my own creation. That said, it is my interest in polling and the explosion of sophisticated techniques that is of particular interest today. This year in particular, we have seen political polling and its complexity (to put it politely) as demonstrating the proof of Benjamin Disraeli’s statement that statistics were a form of deception, especially since statistics have been increasingly used to attempt to sway opinion rather than to merely reflect it.

To be fair, polling is an art rather than a science. Until someone’s votes have been counted, it is not certain that they will vote. Likewise, voting is affected by all kinds of fluky factors like weather, and motivation to vote is stronger in some populations than others, making it difficult to match polls with actual behavior. The obvious importance of determining who is likely to win a given context has made analyzing and unskewing polls a massively popular endeavor. One of the ways this has been done recently is by examining cross-tabs. Because the people who collect polls are often fairly obsessive about information and statistics, it is possible to see aggregate data on the people who make up their polls, which can then be examined and critiqued.

It is at this point where statistics become lies. Based on the assumptions in the data, one gets different results in a poll. For example, music charts drastically changed once data was collected automatically than when it was collected by stores. Previously, albums would rise slowly and peak, and then slowly fall again. Once information was collected in real time, albums tended to debut high and then fall once they were established acts, and only new acts showed the older slow rise and fall as they became familiar to their audience. Did buyer behavior automatically change once data was collected in real time? No, but the charts changed, because the charts reflected the behavior of music buyers rather than store owners. And that made all the difference.

A similar dynamic has led to growing unskewed polls that seek to correct the biases of polls based on their data. Unskewing takes a poll with a suspect ratio of parties or ethnicities or gender and then seeks to rebalance the poll results based on more reasonable estimates. Given a particular breakdown, different levels of turnout ratios lead to different predictions about results and margins of victory. While some people doubt the efficacy of such behavior, it is clear that the proliferation of people examining and critiquing these statistics makes the data of polling even more suspect in the eyes of many and leads to paradoxical behavior. The paradox is that greater statistical sophistication leads to the increased importance in the eyes of many of intuitive and qualitative measures (what kind of campaign someone is running, where people are campaigning), since the data has been made so suspect in the eyes of many.

I think that Benjamin Disraeli would be pleased that the leaders in debunking statistical bias would be people who shared his political worldview, and that sufficient information is known about many statistics to demonstrate that statistics are often a subtle form of mendacity. No doubt the vast majority pollsters are sincere and honest people seeking to capture reality through the aggregation of data, but recognizing the limitations of that data ought to temper the conclusions that are drawn from that data. When the population of a poll is highly unrealistic, we ought to be greatly skeptical of the results of that data, even if unskewing raises a whole host of questions itself (including the rather metastatistical question of how we know with a high degree of confidence what the data would look like unskewed). In the end, we have to live with even higher degrees of uncertainty when we know how statistics lies, and to be fair we ought to realize that political polling is a special case, allowing us to concede the general reliability of mathematics while conceding that it is likely to be least reliable when the stakes are the highest. But statistics is hardly alone in that failing, something that Benjamin Disraeli would have known a fair amount in his own career and life. Such is life in a fallen world.

About nathanalbright

I'm a person with diverse interests who loves to read. If you want to know something about me, just ask.
This entry was posted in History, Music History and tagged , , , , , , . Bookmark the permalink.

2 Responses to Benjamin Disraeli’s Revenge

  1. Pingback: Each And Every One | Edge Induced Cohesion

  2. Pingback: Common Core And The Politics Of Math Education | Edge Induced Cohesion

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s