## Correlations between Campaign Spending and Green Party of Canada vote outcomes.

Now there’s an exciting post title for you! Seriously folks, I have seen a little discussion here and

It Matters

there about just how important election spending is to winning votes. I thought I would sink a little time into quantifying the relationship between EDA formation, campaign spending, and electoral success. I trotted off the the Elections Canada website financial reports database, and the election results database. It’s pretty easy to get sortable data from here, all you have to do is look up the relevant Candidate, Party, or District, and then download the data as a txt file. This file can be imported into an excel spreadsheet, and viola, you have sortable datafiles to play with.
It actually took me a little time, because I had to sort the data so it was all in compatible rows and columns, but I finally got every Electoral district in Ontario, into a spreadsheet, along with every EDA that filed a return in 2007.

The three sets of data I merged were candidate financial return summaries, 2007 EDA financial return summaries, and the actual election result summaries. The data is incomplete, due to some late filings, and of course, not every campaign reports all the data consistently. Still, there were some basic, and irrefutable findings to share. I wish I could figure out how to upload an excel file, then I could put a link here so you could download the actual spreadsheet, but I’ll have to settle for the outcome of some basic statistical analysis.

 Correlation between spending and % Vote: 0.76574 Correlation Between 2007 EDA Assets and %vote: 0.534114 Correlation between Transfers into campaign and % vote: 0.726926 Correlation between campaign contributions and % vote: 0.277441 Correlation between spending and total votes: 0.750952 Correlation between EDA assets and total votes: 0.575364 Correlation between transfers into campaign and total votes: 0.697996 Correlation between campaign contributions and total votes: 0.329326

election-spending

So there it is folks, this table was produced using the basic CORREL function in Excel. A correlation of 1 means a perfect positive correlation between the two variables, and a correlation of 0 means no relationship whatsoever.  As you can see, there is a very strong relationship between reported election expenses and both the total number of Green Party votes, and the percentage of the total vote. The relationship between the financial position of the EDA the year before, and the actual vote outcome is less important, but still pretty strongly positive. Whether the funds were raised by the Campaign, or by the EDA prior to the election was still a positive relationship, but much less influential.

For what it’s worth, the data supports the following conclusions:

1) Raise as much money as possible, from whatever sources you can find.

3) Make sure that your campaign has a finance chair, and continue to raise money throughout the campaign.

All these things will have a strong positive outcome on your vote come EDay. Please don’t bombard me with criticism about the nature of causal relationships. I know that in many respects the money is a symptom of organizational strength. That’s why I ran a number of different correlations. This data is weak in many respects, BUT it does demonstrate that even in the absence of an EDA organization, hard cash still has a major impact.

As to how you spend your money, I wish that EC laid out the detailed spending reports in an easily managed format, because then I would irrefutably prove that local advertising is a waste of F***ing money. If you cannot canvas widely, your advertising should be in the form of flyer’s, widely distributed, which will drive people to your website. Do everything you can think of to ID voters, build your mailing, and emailing lists, and get more supporter data stored away for the next election. Unless you have a lot of eday volunteers to GOTV, you can pay for voicemail message delivery, and telemarketers to GOTV for you. I don’t have any hard data to prove it, but I believe that broadcast recorded message drops into voicemail will be effective in converting voters. Keep it fun, and make sure that you have plenty of chances for volunteers to party a little. These are some of the appropriate uses for all that money you’re going to raise for the next campaign.

As a sort of a post script to this, there is some happy news for 18 Ontario EDA’s. These are the campaigns that broke the 10% barrier, thus being entitled to a rebate on 60% of their election expenses. The rebates will range from a low of \$2,800 to a high of \$46,000, so that means a fair number of teams will go into the next election with a substantial war chest. One the other side of the coin, it really hurts to see Ottawa Centre miss their \$24,946 rebate by a mere 45 votes! There were altogether another 10 campaigns that missed their rebate by a few hundred votes, so that goes to prove that even at our level, every vote counts!

### 23 Responses

1. Whilst I do not disagree with that correlation it is a sad commentry on the state of our demorcracy when the amount of money spent effects the outcome of elections to that extent. A correlation with other partys in this regard would be “interesting”!
http://democracyunderfire.blogspot.com/

2. Great analysis.
To complement this, last year I loaded up past results into my chart-o-matic machine and came up with the following:

http://vote-green.ca/metrics/

I haven’t included 2008 results because they are so damn depressing.. I expect they’re similar.

I propose Ogilvie’s Equation:
V = S + B/C + P + L + HQ
where
S = base support
B = campaign budget
C = cost-per-vote, this year
P = vote bonus/penalty for the personal efforts of candidate
L = vote bonus/penalty for the party leader
HQ = vote bonus/penalty for the efforts of the central campaign team

In 2008, for a typical candidate this worked out to
V= 2000 + \$6,000/\$3 + P – 500 – 500

This works out, on average, to 3,000 votes, which, magically, is the average GPC vote in 2008.

• An interesting equation, reminescent of Keynes general theory, Y=C+I+G+(X-M). The devil of course, lies in the details, and reliance on stochastic measures of past relationships vs. fundamental investigation of the real subjective decision making processes of economic actors. (or in the case of Ogilvie’s Equation, electoral participants). Still, it’s better to start with hard data, and draw what conclusions you may, then crack the tougher nuts.

3. Allow me to post my own contribution in this field, looking at the overall correlations by party by election in federal elections from 1997 to 2006.

http://www.punditsguide.ca/2009/02/patterns-of-candidate-spending-in-past.php

There is not a complete dataset for 2008 as yet, and the correlations there are weaker thus far (probably due to the characteristics of early vs late filers, I’m guessing).

There are some significant within-group differences, however, that I’m planning on publishing something about very soon.

Thank you for publishing this dataset.

• Hi Guide,
I visit your site pretty regularly, and you’ve done a lot of good work. In the case of this dataset, I deliberately focused on Ontario, so as to exclude some of the pretty wild regional variance in GPC electoral outcomes. Ontario is relatively homogeneous(yeah there are regional discrepencies, but not so much as Canada wide). Ontario is a large enough sample for a degree of significance, and really, I was trying to isolate the raw impact of \$., and determinbe a rough proxy for EDA organizations impact on electoral outcomes. I felt guilty not throwing up the raw data, because there were quite a few late filers, for which I have the vote outcome, but set the spending and transfers at \$0. I guess that since my contention was that spending really matters, and this bastardization of the data mitigates strongly against my conclusion, it bothered me less because of my conclusions than because the data wasn’t as rigorous as I would like. The long and the short though, is that when the final data comes in, a repeat of this analysis will reveal an even stronger correlation than shown in this post.

4. To be sure, and just to emphasize that my comments were not in any way a criticism of your work, BGB, merely an explanation for why I had not extended my own look from 2006 out to 2008 as yet.

Indeed the data in my blogpost cited above predate my most recent review and update of the 2004 and 2006 data anyways. But I have a few other commitments to get out of the way first, before I can get back to further work on this front.

5. You’re probably better off to wait a month or so until all the late filers who are going to file, do file. If I did it again, I would delete all data from non-filers. The really thorny issue is to segment data for regional variations in support levels that are not related to spending, adjust data, and run the anova to isolate the impact of spending on electoral outcomes.
Geez, I guess you could easily make a life’s work out of crunching this data. It’s a pretty important question though to answer HOW MUCH does money matter in politics? Answer that question credibly, and maybe further election finance reform can be talked about seriously.

6. I posted this on Jurist’s blog when I could not get to this blog. I’m not going to even try going to the EC site until later.

Here is what os essentially a sample size consideration.

Very few EDAs get significant contributions on their own. Across the board.

In the GPC I think the only exceptions to that would be Chris Tindal’s byelection campaign [possibly Daniel Grice as well, never looked at that].

And then the very big exception of Central Nova. But that is really a parallel national fundraising operation that “happens” to be run out of an EDA.

I’m going to have to wait to look at the numbers [currently on an old computer in a lobby in the middle of a snow storm].

So I can’t see what numbers you used BGB, but you mentioned ‘anecdotally’ that only Guelph [plus obviously Central Nova and Vancouver Centre] got transfers.

My spies tell me there actually were a lot of such transfers to campaigns last Fall- many without being asked for or terms mentioned. [Called “loans” after the election I am told.]

And then there is GPC revenue sharing of subsidy monies with the EDAs. I know this had been “suspended” before May came in, but implicitly I understand it was to be re-inststaed.

Been meaning to ask about that. If there is/was revenue sharing it would show as transfers.

7. Based on my experience, I would expect the substantially smaller correlation between votes received and “longer run organizational health” indicators such as 2007 EDA balance.

Spend more, get more votes seems to be by far the main factor. [With the caveat that on the aggrgate parties are going to allocate spending not optimally, but where there are visible expectations of success.]

And this, bluntly speaking, is where I beleive things like 2007 EDA balance come in: An EDA with a balaance is more likley to be deemed worthy of the resources that can be bestowed by the central campaign…. which in turn are the actual drivers of likelihood to increase vote share.

8. Hi Ken,
It’s a bit of a pain in the ass to correlate and track all transfers and refine proxy measures, by virtue of the inadequacies of EC’s data. When the 2008 EDA financial statements go public, then we will have a 100% supported picture of what happened.
Revenue sharing is not with campaigns. It is with organized EDA’s that maintain an executive, and meet various organizational criteria, like regular EC filings, timely candidate selection, full slate of EDA Exec’s. In the past it is suspended until election debts are repaid, don’t know if that’ll be the case this time. (I’d be surprised if it weren’t). For the 2007 EDA Financials, you can see that the revenue sharing transfers in were a pretty small proportion of most EDA’s assets. In addition, a proportion of the funds raised by some of the EDA’s, are processed by the central Party, then transferred to the EDA by direction of the donor. The upshot is that the Transfer numbers are, and will remain cloudy for all time.
I think that given the quality of data available, the year end assets of each EDA aren’t a bad proxy for EDA organization. When we get the 2008 numbers, I’ll rerun these sets, with a few more variables. I have seen some interesting comments here, and I would like to have a stab at quantifying some of these relationships a little more rigorously. The data will always be limited though, and without big\$\$ nobody is going to be able to precisely quantify the relationship between money and electoral outcomes.
There were three loans recorded to Ontario EDA’s. the terms, and identity of the lender isn’t disclosed in this dataset. I think that your’ ‘spies’ probably have a parochial view of the GPC, and maybe don’t interact with many of the nationally quiet, but locally active EDA’s. ‘a lot of such transfers’ could mean a dozen nationally? That’s not a lot in my book. Especially when we’re talking about a couple of thousand bucks. There are quite a few EDA’s that chug along, doing their own thing, whilst ignoring everything at the hub, except the occasional revenue sharing cheque. These ones form the backbone of the data.

9. It would be interesting to do some kind of value for money comparison between on the one hand spreading tranferred funds around EDAs broadly and thinly, versus funds expended on campaigns where the GPC has thrown in the kitchen sink to win a seat.

So how many vote increases per thousand dollars of expenditure for each pole of the comparison.

In the two years between Mays run in London North Centre and the 2008 general there has been those 2 campaigns of hers, plus Carr’s Van Centre campaign, and the Guelph campaign [the by-election that got trumped by the general].

105K LNC
55K CN campaign [reported]
23 CN campaign salaries paid by GPC
80 CN spending before election
94 Van Centre
160 Guelph

Carr’s salary arguably belongs in that column also. A party has to pay a Leader. The logic of ALSO paying Carr’s salary has a great deal to do with getting her elected- give up on the priority of electing her and that salary becomes a luxury.

While all those campaigns [except LNC] did fundraise, a huge proportion of that was outside their ridings… essentially from the same pool where funds come from that would be transfers broadly to many EDAs.

And there is about \$15,000 per month that goes into maintaining staff and the permanent headquarters in Central Nova.

I don’t think there is much doubt that the cost per vote increase of those high profile campaigns is many times higher than the cost per vote of spreading funds broadly among EDAs.

Of course, the ‘value’ of the GPC actually electing an MP cannot be bean counted. In principle, if it cost \$20 per vote, that might well be worth it.

But since there is no certainty at all that throwing in everything to campaigns- much less to a few of them every election [and Mays even between elections]….

…. then comparing the cost per vote increase for the two ‘methods’ of party building is a valid one.

10. In the four ridings with top GPC spending the before and after vote gain was 29,500.

The aggregate spending in the list above is \$517,000.

That works out to a “value for money” cost of \$17 per vote gained.

If anything, that is somewhat understated. Carr’s salary was not included, nor anything for the obvious non-reporting of GPC paid salaries in the Guelph and VanCentre campaigns.

Looking to the future, about \$20,000 per month has continued to be spent since the election in Central Nova and on Carr’s salary. A year of that and even only two full tilt campaigns in the next general comes to another \$400,000 investment [more if the general is later than this Fall].

And that additional \$400,000 will bring in a much smaller vote increase, not to mention a likely erosion in Guelph without a full tilt campaign there.

========

Once the 2008 candidate campaign and EDA filings with EC are complete, I don’t think it would be difficult to aggregate the numbers for a per vote costing of ‘broad and thinner’ transfers to campaigns.

When you look at the candidate campaign or EDA filings it is easy to tell actual transfers from the GPC from transfers that are ‘flow throughs’ of contributions made to the GPC for a riding.

Perhaps that cannot be done automatically. But there is a lot of good reason to be selective in the first place about which campaigns to look at.

A sample of campaigns that is representative would give more meaningful results than an aggregated mess that includes a lot of ‘campaigns’ with no organization, etc.

11. Interesting thoughts ken. I think that truly accurate analysis would be tough to acheive, but I bet I could construct a very convincing case for just about any strategy by careful data selection.
I have spent hundreds of hours on the phones in the past talking with EDA exec’s across the country. Without any insult intended to the EDA’s, many of them are woefully unprepared to effectively spend money during a campaign. I promise you that if every election penny were spread around every campaign, then about 50% would be either unspent, poorly spent, or outright wasted. The primary objective of the GPC must be to elect MP’s, and this must happen in some specific riding somewhere. Maxing national votes won’t cut it long run. I think that an intelligent targetting of key ridings in every region, and concentrating efforts on these EDA’s will have the best combination of return on invested \$, and closing the organizational gap to the point where normal, high quality candidates will be attracted to the campaigns. Perhaps spreading the money a little more thickly between 40 EDA’s, while assisting a basic level of organization across the 308 ED’s?
The Guelph spending was a good bet due to the by-election. Also your comment that the braod fundraising cannibalised funds that would have gone to other EDA’s isn’t true. That’s not the way donor’s work. An appeal was made for a specific and compelling case. It worked. This is precisely how the CPC maximises their fundraising efforts. By repeated specific purpose appeals. Their calls to action are always related to a specific policy outcome, or advertising campaign, or some such. It works, and doesn’t really cannibalize much.

12. Obviously the goal is to elect MPs.

Maybe its best to eliminate straw persons first.

I would argue that the status quo of spending a few hundred thousand per year on the ‘spend till we get there’ “strategy” is not representative of what judicious high spending in particular ridings could be.

But it is also a straw person for you to suggest that spending ‘broadly and thinly’ means sending money to every campaign whether they are ready or not.

And, yes, anybody can manipulate data sets to prove whatever they want. But it is also true that people can make good faith efforts to take data that is very limited in how rigorous it can be used, and use it to make better choices.

When the GPC central campaign distributed transfers to candidate campaigns last fall, no doubt they made a number of dubious choices.

But it would still be instructive to see the ‘value for money’ voter outcomes of those transfers.

Its also questionable whether the staus quo counts as ‘targeted spending’. Even if Guelph was a good choice… its still part of a package of more than one place of throwing in everything possible. So its valid to sit back, measure the cost and figure whether its a better way of getting to the goal of electing MPs.

Increasing the vote share in a lot of ridings is not just a numbers game, or a game of increasing the public subsidy per vote funding. For some of those EDAs that see their vote share go up, it will be the start of lasting growth.

You point about targetted fundraising is correct- and it is not a zero sum game. But from what I have seen looking at the sources of revenues for these campaigns, I would peg at least 2/3 of it coming from general revenues [be that subisdies or donors] rather than targetted fundraising pitches.

13. Apropos of the GENERAL topic…

The GPC brain trust has for some time clearly attatched a high enough priority to spending as much as possible in particular campaigns…. to the degree of substantially stretching the spending limits set out by the act.

This has been done carefully and deliberately- presumably to sound out where the ‘letter of the law’ boundary is as to how much “stretching” Elections Canada will tolerate.

Quite a bit apparently. To the tune of about 20% in London North Centre, which was reviewed and accepted by EC.

On paper at least, the “stretches” of spending limits have since then been somewhat more modest, but with generally more apparent plain ommissions of expenditures.

Leaving aside questions of legality- there is the meaning of ‘doing politics differently’. I personally feel this rivals how the Harper crew plays fast and loose with the spirit, if not the letter, of our system of electoral financing regulation.

But leave that point aside too. For the purposes of discussion, lets call the practice of extreme stretching of riding campaign spending limits ‘fair game’.

And as such, it does show just how far, and in how many cases the GPC brain trust is wedded to ‘make or break’ campaigns.

Following this stuff is just too geeky for words. So I’ve decided, for now at least, not to write up a “how it is done” guide to extreme pushing of spending limits.

But anyone interested in a verbal summary can send me an email kenpat@ns.sympatico.ca

14. Long time reader, first time responder.

Allow me to introduce you to Anscombe’s quartet: http://en.wikipedia.org/wiki/Anscombe%27s_quartet

It is not sufficient to have correlation. What would really be good would be to see the data graphed. In particular I note that even with a correlation of 1, we could have infinite positive slopes for the regression line. Thus, the 8 items that have a positive correlation may not be as powerful as suggested.

For example, with the same correlation between spending and % of vote, I could have slopes of 0.1 or 0.5 or 20. That is to say, for \$1 more spending I could expect 0.1 more votes or 0.5 more votes or 20 morew votes! That’s a big difference!! So, looking at the slope of the regression lines is vital. Can you send that data my way? I can recrunch it if you’d like and put that new math teacher degree thing to good use.

• Cool!
The data is a little difficult, because when i grabbed what was available, a fair number of EDA’s and campaigns had not yet reported their expenses. I’ll dig up the original spreadsheet when i get home, and email it to you. If you’re willing to prepare something a little broader, then I’ll put in the couple hours work to pull all of Canada, and present it consistently. I guess that’ll have a problem inasmuch as regional variations in underlying data will be subsumed by the ‘global’ dataset, so we can get a true national picture of votes per marginal dollar, but we’d be describing the variability between regions as much as the marginal value of a dollar spent. Still, the more data the better, and I’d love to have a well supported number.

15. That would be awesome to have all the data, too!

The four causes of higher Green turnout/raw votes you hypothesize are interesting. As a campaign manager in lonely Elmwood-Transcona (I’m not from that riding and neither was the candidate – and we received 0 hours of support from the local Greens….barely plural), I would also love to see the correlations of votes with (1) memberships in an area; (2) number of official EDA executive positions filled; (3) number of active volunteers; (4) number of volunteer hours clocked; (5) spending in the years before an election; etc…

But, that data set is going to be harder to come by. (1) and (2) are possible, though.

I’ll also be able to seperate out the provinces like Manitoba which are outliers; and I can eliminate the non-EDA ridings to check and see how much noise that causes.

Anyway, send data. Will compute.

• I would love to see all those correlations as well. 1 is possible, 2 is not as relevant, because as often as not the position may be filled, but the officer isn’t always active. I suspect a better variable is simply, does an EDA exist, and for how long?3 is problematic, because it’s a normative judgement call, and 4 is highly improbable to have been tracked, except by a few anal characters like myself who measure everything. 5 is available, but is really a labour of love to assemble the data into some kind of database, (probably an excel csv, or spreadsheet). I’ll do some digging tonight, and email you later.

16. 2 is relevant to the extent that if you have an EDA with a single officer versus an EDA with a serious roster of officers, one would expect a difference in results…but, you are right that activity is more important than existence. This is the problem with many memberships – they are members in name only.

17. BGB,

Did you have a chance to find that data yet? If not, you can always just send the Ontario data you used to find the correlations and I can use that data.

(I assume wordpress also sends you my email address, right?)

18. […] Posted on June 20, 2009 by bluegreenblogger In March, I published a post on the correlation between campaign spending and vote outcomes for the Green Party of Canada. I gathered together the Green Party of Canada Campaign spending data […]

19. […] In the 2008 election, there were 41 Ridings where the GPC candidate exceeded the 10% threshold whereby they received a 60% rebate of their election expenses. Probably a paramount and overriding campaign objective  for them, is to repeat this feat in the current election. Ditto for the host of EDA’s that were within a percent or two of the 10% threshold. Chances are, if you were in this category in 2008, you don’t need me to tell you what to do. While that hasn’t stopped me before, for now I will stick to reminding you that the advanced polls are of vital interest to you. The Green Party Vote will once again soften, and move to other Party’s in the final days of the election, so it is vitally important that you get your vote out in a big way for the advance polls. Put them in the bank before they change their minds, so to speak. In the last General Election, Ottawa Centre missed getting their \$24,946 rebate back by some 45 odd votes. Ouch! How much you wanna bet that this time they will have a strong GOTV to the advance polls? lol The practical difficulty for these stronger campaigns will be that there will not be much of an Air War to support their campaign, and believe me, that will have a disproportionate impact on the stronger ridings. Ridings which have built their support to the 10% range will have to reach out to a much higher proportion of their electorate to win more votes. Presumably, the bulk of the motivated Green voters in these ridings are in fact already being mobilized, so reaching more and more undecideds with a conversion message is their route to improving their standing at the polls. Without an Air War reaching masses of people with a simple repetetive Campaign message, it will be hard slogging indeed to make fresh inroads. I wish them luck, as the only antidote is to pour it on at the local level, which will take plenty of folding money. If you doubt that money is important, have a look at some of the simple analysis I did after 2008. (Correlations between Campaign Spending  and Vote outcomes). […]