Causation and correlation


In response to Stephen’s post, The Mischief Rule, Michael points in a comment (by the way thanks for reading and commenting)  to then Senator Obama’s victory over Senator McCain in the 2008 US Presidential election and suggests that Obama’s ability to outspend his rival was one of the factors that led to his victory:

There were no doubt many factors that contributed to senator obama defeating senator mccain but i think it would be hard to maintain that his ability to massively outspend his opponent on the ground and in the air in all the major swing states wasnt one of them.

In the case of Obama there was certainly a correlation between outspending his opponent and electoral success but was there causation?  I doubt it.  The Republican party was relatively unpopular in 2008 and McCain was always going to face an uphill battle.

As for Obama’s ability to raise large amounts of money, Stephen has already noted that Obama’s ability to persuade voters to vote for him was correlated with his ability to induce campaign contributions. I agree. It’s not surprisingly that candidates that convince people to vote for them find it easy to convince voters to donate to them as well.  I also suspect that donors find it more rewarding to donate to candidates that they think are likely to win as opposed to candidates who they think are unlikely to win.  Jesse notes that political fundraisers “encourage donors to feel that they are part of a community”.  People generally prefer to be part of winning communities rather than losing ones.

The studies into the link between campaign expenditure and electoral outcomes do not provide a particularly strong basis for concluding that electoral outcomes are highly dependent on campaign spending.  Jesse, Stephen, Yogesh and I canvassed some of those studies in our submission on the Ministry of Justice’s Issues Paper which you can find here (the discussion is at paras 8.5-8.17).  A study by Steven Levitt is particularly interesting. I’ve reproduced below the summary of it that we included in our submission:

Studies which seek to isolate the effects of increases in campaign spending that are unrelated to a candidate‘s direct appeal to voters are one way to determine the nature of the relationship between campaign expenditure and election outcomes. While these studies are difficult to undertake, some recent studies have used sample selection to reduce the bias shown above. One study sought to isolate campaign spending by limiting the sample to those elections in which the same candidates faced one another on multiple occasions – a sample which consisted of 633 elections in the United States between 1972 and 1990. Assuming that candidate quality therefore remained the same, and controlling for other factors such as incumbency and national-level partisan swings, the author found that “campaign spending has an extremely small impact on election outcomes”.

I acknowledge that there are of course other studies (and we referred to some of those in the submission).

There is another point.  Not only is there good evidence which suggests that campaign spending is not causative of electoral success, even the correlation between campaign spending and electoral outcomes is not clear cut.  Consider the table below (also from our submission to the MOJ) which compares spending in the 2001 and 2005 UK general elections.

Despite spending less in 2001 than the Conservative Party, the Labour Party won far more seats. However, despite increasing its spending by nearly 64% in the 2005 UK parliamentary elections – and despite spending virtually the same amount of money as the Conservative Party in 2005 – the Labour Party had a significant net loss of seats. It is possible to surmise a number of reasons for this based on, for example, national partisan swings or the likelihood that many Labour Party gains in 2001 were in electorates which would be difficult for that party to hold in the long-term. However, there seems to be no basis for explaining the changes in electoral performance based on spending.

A comparison of the 2005 and 2008 general elections in New Zealand is similarly inconclusive.



10 Responses to “Causation and correlation”

  1. Michael Appleton Says:

    Jono, hopefully my comment below on Stephen cover off some of your points.

    I think we’re probably arguing over a small area of ground here, given that I acknowledge that: 1) A minority – sometimes a very small minority – of voters is persuadable by an election campaign; 2) The ability to spend a lot of money is only one of many factors that influence this pool of persuadable voters (spending *quality* is of course very important too – Obama’s better targetting of his resources towards delegate states, for example, was an important factor in explaining his win over Clinton in the primaries).

    Given these concessions, the question then becomes: is this influence of money, restricted by the above, worth regulating? I think it is – probably because the consequent freedom of speech restrictions don’t concern me as much as I suspect they do you.

  2. Jonathan Says:

    Thanks Michael. I agree with you that spending quality is likely to be a much greater factor than spending quantity.

    I also suspect that you are right that much of our disagreement here largely revolves around how much we each of us is concerned by the restrictions on freedom of speech involved.

    I suspect another reason for the disagreement, however, is that I am inclined to think that the amount of money a party or candidate can raise is likely to reflect factors such as the resonance of the party or candidate’s message, popular support for the platform and an assessment by donors the party/candidate’s chances of success. I take it from your response to Stephen below that you think the amount of money a candidate/party can raise is likely to be more determined by the underlying wealth of that candidate/party’s core supporters?

    One reason I am doubtful about that is that I am unconvinced that income/wealth and political ideology are strongly correlated, at least not among people who are likely to donate to politicians.

    Finally, I’ve been thinking about your point about why politicians keep spending money if it isn’t very effective. I’ve got some thoughts about this which I might blog in a separate post later but to explain briefly here, I think the most likely reason is that the marginal cost of spending more on advertising is very low. Given that in most countries politicians cannot spend campaign funds for their own purposes if they are not exhausted in a campaign, the opportunity cost of spending more is quite low. The only real cost is the cost of raising the money (which I think is measured by the candidate in terms of the amount of time spent at fundraisers, costs of running websites etc). If a popular candidate is raising a lot of money, particularly on the internet, then it makes sense to keep spending even if that will only have a minimal effect. Also, given that most elections operate on a winner takes all basis, there are high incentives not to leave money in the bank.

  3. Jonathan Says:

    Also Michael, thanks for reading and commenting 🙂 I hope to post about something other than campaign finance reform this week. I’ll try find an area where we are in agreement!

  4. Michael Appleton Says:

    Jonathan: I think both explanations for fundraising success have merit – that is, are part of the overall picture: ie does the candidate have a message that resonates with the public and can the candidate present a message that people/organisations with money to donate are comfortable with? Both of these are important questions in determining fundraising success. On economic issues, there is probably a limit to how left you can go before corporations are simply not willing to open their chequebooks to you. (The same is probably true on moral issues: go too conservative in New Zealand, and certain donors would be scared off.) As an example: while I’d need to go back and check, I’m fairly confident in asserting that Act has been more successful in fundraising than the Greens/the Alliance in our five MMP election campaigns to date.

    I agree that there is a strong incentive for politicians to spend money, once raised. But if you read any campaign books, you’re struck by how much candidates hate fundraising. So why keep doing the fundraising itself, if you could just as profitably (or more profitably) spend the time doing media interviews, giving speeches, shaking hands or even formulate policy?

  5. stephenwhittington Says:

    I’ve gone back and compiled the figures.

    The Alliance spent more than the ACT Party in 1999 (and if you add the Greens expenditure in, about 50 percent more). The ACT Party spent a lot more than both in 2002, more in 2005. The Greens spent more than ACT at the last election.

    So, for all MMP elections, it’s 3 – 2 to the ACT Party. In politics, the score is 24 – 16 to the Green Party (MPs).

  6. Michael Appleton Says:

    How interesting.

    The figures are, if I am reading the tables correctly:
    1996: Act $1.7m; Alliance $558K
    1999: Act: $658K; Alliance $745K + Greens $236K
    2002: Act $1.6m; Green $598K
    2005: Act $1.4m; Green $835K
    2008: Act: $1.4m; Green $1.7m.
    Total: Act $6.8m; Green/Alliance $4.5m

    So, Act’s spending is fairly consistent, except for 1999, whereas the Greens’ ability to fundraise has improved over time…

    I’m not sure who this helps, in terms of our disagreement. On the one hand, as you observe, the Greens have been able to get more MPs elected on less money. On the other hand, part of Jono’s thesis is that money follows political support/resonance with the public. If that were completely true, wouldn’t you expect a more successful party to attract more money? (Weird also that 2002, the year when the issue of GE food was at its most potent, attracted so much less money than 2008, when the chances of the Greens having a role in government seemed very small.)

  7. Alex Says:

    I don’t like the methodology discussed in the bottom part of the blog post. It seems like you either compare spending between parties or on the same party accross time. You note that the party that spent the most didn’t always get the most votes or that increasing spending over time may not gain a party more votes.

    The problem is that many other factors likely have far stronger effects on voting outcomes. This is not to say that spending doesn’t have an effect at all.

    What you need to try and do is figure out for one party in one election what the impact on votes would be if that party spent less in that election. i.e. Jonathon notes that UK Labour increased its spending by 64% between 2001 and 2005 but won fewer seats in 2005. That’s well and good. The question that needs to be answered if even fewer seats would have been won if UK Labour had only increased its spending by say 34%. The comparisons in the bottom of the blog post don’t help answer this question. I’m sure that other academic articles have attempted this but I am unfamiliar with the literature.

  8. stephenwhittington Says:

    There are such articles – a few are referenced in our submission, and two are referenced in my post on mischief below.

    Michael, the point isn’t that success = votes + money, but that there will tend to be a relationship between electoral success and money, but that this can be partially caused by another variable – how well you convince others. This relationship is overlaid with a whole lot of other aspect, like strength of belief, etc.

    A possible reason why the Green Party may do better in 2008 in terms of funding is because it provides a better signal of true belief, and therefore increases value from consumption – “Yes, I know it’s a waste of money as they won’t get in, but I believe it so firmly that I will give.”

    As a personal example, if The Seasteading Institute were popular, I probably would not donate.

  9. Spending in the New Jersey gubernatorial race « Today's Dissent Says:

    […] in the New Jersey gubernatorial race By Jesse Jono has previously posted on the uncertain relationship between election spending and election outcomes (noting that other […]

  10. il health and beauty natural oils Says:

    However it’s a slippery slope from attempting to eat proper
    to creating an consuming disorder.

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 )

Google+ photo

You are commenting using your Google+ 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

%d bloggers like this: