Everyone knows that Obama’s Democrats just got their arses kicked in the mid-terms. The dominant political narrative is that it was a disaster, right? Those stupid Dems, acting on political principle and bringing in a second rate universal healthcare system by running roughshod over the bipartisan Republicans to overrule democracy, no wonder they lost the mid-terms right? This is the dominant narrative amongst political pundits, and all debate seems to be focused on Obama’s repudiation, the coming destruction of Democrats, and the fundamental conservatism of the American electorate. This is how pundits speak.

I’m not American, so although I feel myself qualified to speak to some extent on the theory of American healthcare policy, I don’t feel like I really know anything about US political debates, so I just assumed that the story being presented by these pundits was some kind of biased but basically correct analysis of the situation, I think you know the kind of interpretation I mean, regardless of your politics: accepting the broad truth of the facts presented but maybe having a different interpretation of the implications and the long-term trends and outcomes for your preferred side of the debate. As a non-American who thinks Obama is too right-wing to make it in my own political culture, this is all just academic angels-on-heads-of-pins stuff. However, recently I stumbled on a counter-narrative to the standard view of Obama’s bloody nose, and it got me thinking about the broader issue of how political pundits analyse politics, and whether anything they say is worth anything while they remain ignorant of statistics.

The counter-narrative is that Obama did better than expected, and although the Democrats got a bloody nose in the elections this is normal, and they didn’t do badly at all. This counter-narrative is exemplified by this post I found on Greg Laden’s blog, which is so simple it’s ridiculous. Laden plotted number of seats lost vs. presidential approval rating for all mid-term elections since 1946, using Excel, and showed the line of best fit. The implications are pretty obvious – there is a direct relationship between presidential popularity and number of seats lost in a mid-term, but this mid-term election Barack Obama was not at the bottom of the range of approval ratings, and he did significantly better than presidents would usually do given his unpopularity. The Democrats should have lost about 40 seats in the House, but they actually only lost 10 – a huge victory. A weaker analysis of the Senate suggests he was facing a uniquely hostile senate election, and probably did well given the circumstances. These analyses are very rough and could be done better (e.g. logistic regression adjusted for number of seats, confounders for presidential party etc) but they probably wouldn’t improve on the basic findings: the Democrats actually did uniquely well given their circumstances.

My guess is if you extended these analyses to an analysis of results of the following presidential and congressional elections, you would find that the number of seats lost in the previous mid-term was a big predictor of success, with the implication that even though most political pundits are saying Obama has flumphed it, the data implies Clinton’s job is going to be extra easy thanks to the Big O, and the congressional elections will also be better than normal. But I suspect my reader(s) have not got that impression from the popular press. Where have pundits gone wrong?

If one thinks about the average pundit, it’s pretty obvious what the problem is: they’re a journalist, and journalists are generally thick as shit. Your average political pundit is going to be barely literate and completely innumerate, and they are going to be firmly of the belief that their experience, connections, colleagues and unique insight into the way people in the halls of power think gives them a unique power to analyze political events. I think this is not true, and that it’s physically impossible for modern political pundits to have a broad enough base of experience to provide any insight into the political process: instead, they are simply retailing gossip from their mates. I will prove this by comparing political pundits to actually intelligent sports journalists, and for my example I will choose that most reviled of sports commentators, the UK football (soccer) pundit.

Let us first consider the average political pundit in the USA. He’s a man, obviously (this is serious stuff!), aged in his 40s, with a 20 year career. Given his age he probably graduated from a journalism degree (they were starting to take off 20 years ago) but he may be a graduate of some other field. He probably has some “Econ101” (in the derogatory sense), so he’s swallowed all the typical rubbish about government debt and how the Chinese own America and so on. He reports deficit busses as if they’re serious politics instead of theatre. Of course he’s an idiot. But he has experience! He has 20 years of experience! But what does this 20 years of experience mean, actually? He has seen, at most, 11 congressional elections (they happen every two years). Of these, if he’s really lucky, three might have been elections occurring in the second term of an incumbent president (e.g. Obama and Bush junior) and maybe six are mid-term elections. Precisely one occurred in the aftermath of a major terrorist event. No one he knows or has ever met has seen a black man elected president. At most 3 of those 11 elections happened against a backdrop of internet-based political activism. He has seen a maximum of six presidential elections, probably four, and has seen the presidency change parties at most three times. He has colleagues that he might be able to discuss elections and the politics with, but these colleagues have seen no more federal congressional elections or presidential elections than he has – they all talk about the same 11 events. This is his “experience.” His boss – let’s imagine a man 20 years his senior – might have 20 congressional elections, 4-5 elections in the second term of a presidency, and maybe 10 mid-terms. This boss has had the remarkable experience of seeing a presidency retained by the incumbent party on a change of president – once. Once. The entire cadre of political journalists have this experience to their name. When it comes to second-term congressional elections, our putative political “expert” is like Lieutenant Gorman in Aliens. “How many drops have you done, Lieutenant?” “Three. Simulated.” What does Private Hudson think of that?

Now let us compare with an equivalent British football pundit. He has seen 20 FA Cups, and assuming he has done his job assiduously has watched every team in the Premier league play every other team 20 times every year. Those games he couldn’t watch he has had a chance to review when needed (eg before crucial matches), giving him experience of watching any one team play probably 15 or 20 times a year, for 20 years. Obviously he focuses on the top teams, but he is able to replay hundreds of games for any team that enters the FA Cup. He has only seen 5 world cups, but he has also seen 5 European cups, and watched the players from the European cup in the Champions League, plus been able to watch international friendlies, African cups, etc. Where he is unfamiliar with a part of the football world, he will have access to mulitple colleagues who are familiar with other areas (Europe, Asia, lower divisions, Americas, etc.) and where he has not been following a team he will have access to a huge compendium of commentary from those who have. There is no chance at all that his colleagues have been raised and trained on the same football experience as him, so when they meet to discuss the football world they will be able to bring potentially radically differing views and experiences to the meeting and, if they take the job seriously, will be able to take away as many different views as there were people at the table. This pundit’s boss will have seen 40 FA Cups, has seen football powers come and go, was around before the era of the oligarchs, etc. This is why football critics are able to identify whole movements in football (like tiki taka) and also identify when they fade away. This is why football pundits can speak authoritatively about, for example, the transitions in the German national team.

The football pundit has experience, and collegiate connections. He or she has access to a broad and deep pool of knowledge. The American political pundit, on the other hand, has hearsay and gossip. The American political pundit cannot claim to have anything that rises to the level of “experience.” He and his mates have seen the same five presidential elections, the same 11 congressional elections, the same two second-term mid-terms. They go to the same cocktail parties and talk to the same media reps. They are peddling nothing better than the perceived wisdom of their five mates, and that wisdom is shallower than the gene pool that your average journalist is drawn from. They are in every way inferior to football commentators.

This problem of political punditry could be solved by application of a simple skill – basic data analysis. Greg Laden’s posts show that, in the absence of experience, even basic data analysis skills can be enormously useful. But political pundits are ignorant of anything resembling data, and they will push back strongly against the idea that analysis can improve their work. As a classic example of this, consider the hostility Nate Silver received for his work predicting American elections, or the way the Poll Bludger was treated in Australia (outed by a major newspaper for disagreeing with their pundits). Because the problem is simple – 5 or 10 events is not sufficient to provide anyone with real experience, at least in a world as complex as politics. It’s not that these commentators are incapable of analysing from experience – they just don’t have the experience. In the modern world, we would never take dating advice from someone with 5 conquests, but we routinely listen to men and women who have seen no more than 5 or 10 elections. This is the level of experience sensible people dismiss as “anecdote.” And in such a circumstance – when you or I know we have insufficient experience to inform our decision – what do we do? We go to data. But political pundits can’t go to data, because they are numerically illiterate. So instead they basically peddle gossip.

The problem with idiots peddling gossip from a politically and socially privileged pulpit is twofold. The first is obvious: they’re wrong. Why should I listen to someone with 5 presidential elections under his belt? How can he possibly have enough knowledge of what goes on on the federal stage to be able to provide cogent commentary? But the second is more insidious: they’re vulnerable to manipulation by powerful or moneyed forces. There are lots of thinktanks out there in the wilderness of American politics, peddling lies that are paid for by various big companies. Why would someone who has only ever seen five presidential elections, and who doesn’t have any ability to properly assess the history of elections, be able to properly analyze political theories put forward by bought-and-paid operatives of the big think tanks? How is experience going to protect someone against the lies these organizations put forth, when their experience is so limited it provides them no context for judging any analysis?

This is an example of the importance of data analysis to modern life. There are many problems where we as individuals lack anything resembling a coherent required level of experience, but where data contains the knowledge of multiple generations, there for us to analyze. Even simple data analyses will take us somewhere, help us to get a context for the problems we’re trying to understand. Someone with hundreds of football games and 20 FA Cups under their belt may be wrong about their understanding of the subtle forces at work in football – they may need data analysis to help them see clearly what is going on in there – but they have the experience to draw from, and indeed if we did data analysis we might well be just looking for numerical patterns in the information the pundit already knows. This is an example in which data analysis can help an experienced expert to better understand their work. But in the case of political punditry it’s different. There is no expert, because people die before they can become experts. Instead we have a gaggle of gossip-mongers. In this case data analysis doesn’t complement their experience, because what they have is not experience, but a couple of moments. Data analysis should replace their work. They are irrelevant, and probably rather than analyzing the actual history of political movements in America, they are regurgitating some canned idea from a think tank. And how can they do anything else? They, personally, are ignorant and naive, but they believe themselves to be experts.

Which is why you shouldn’t believe anything political pundits tell you about elections. At best they’re inexperienced; at worst they’re inexperienced, ignorant, and retelling hearsay from someone who probably has a vested interest in crafting hearsay. Ignore them, and look to the data.