Just read a great post on DemocraticSPACE.com

What does the Data mean?

What does the Data mean?

I just read a great post on DemocraticSPACE.com. Greg Morrow has constructed a matrix to do comparative analysis and ranking of  electoral potential for the Green Party of Canada. What a novel thought for the National Campaign, data driven decision making!  Greg has constructed a matrix whereby he ranks every electoral district in the country by relative standings for the following variables: Membership, votes won, gap between GPC and the winner, numbers of ID’d voters, and then compiles them into a ranking from strongest to weakest riding.

 Normally I have criticisms to make about anything that anybody has to say. Call it a personality flaw, or call it attention to detail  if you’re more diplomatic. In a previous post I discussed the importance of how the vote split falls in determining the electoral opportunities facing an electoral district. Greg’s matrix should probably include a measure of close 2-way, 3-way, and 4-way splits as well. The closer the split, the more different strategies, and opportunities will exist to win votes from other Party’s. For example, if there is a close 3-way split in the vote, and there’s a 20 point spread between the GPC and the first place, the GPC has the opportunity to win votes from each of the three front runners, and build up the 20 point deficit. If there is a way out there front runner, with a 16% spread between the GPC and the winner, then the chances are the deficit needs to be made up entirely at the expense of an incumbent. The incumbency factor is very difficult to overcome, and as far as the GPC annihilating a single Party, if you believe that possible, then I’ve got a bridge for sale you should really think about.

The only other significant criticism I could make would be regarding the ID’d voter data. I can only assume  that Greg has access to civiCRM, and/or GRIMES voter ID data. Many EDA’s don’t use them. I know at least two EDA’s that have significant numbers of ID’d voters resident on custom local d-base apps. Plus there’s the stacks of paper records in most campaign managers basements, waiting for somebody to do the data entry. (Unfortunately, the data entry seldom ever happens, and the data is lost).  In London North Centre by-election, I compiled data on over 9,000 identified Green Party supporters in order to quantify the outcomes from the GOTV effort. The scrutineers sheets were sent to the shredder instead of putting in the two hours of work needed to scan in the actual voters names, so the analysis could never be done, but surely the GPC retained the lists of ID’d voters?

The above paragraph illustrates a point that most professional users of data are aware of. When you formally start quantifying things, run sophisticated analysis, and present it neatly, the conclusions take on a life of their own. They become ‘THE NUMBERS’, and are inherently believable. You need to retain your critical faculties, and dig deeper into the underlying assumptions, and data quality. In conclusion, Greg Morrow’s matrix is a most excellent tool for targeting resources, and ranking Electoral districts. There are a few flaws and/or problems, but they are readily overcome. Be sure to check it out if you plan on holding an opinion on targeted ridings.

Add to: Facebook | Digg | Del.icio.us | Stumbleupon | Reddit | Blinklist | Twitter | Technorati | Furl | Newsvine


29 Responses

  1. Hi BGB,

    I talked about Greg’s post today too: http://www.davebagler.ca/badandgood

    To be fair the CC has been using lots of internal polling to make their decisions. Combine the ability to use excel pivot tables with no campaign experience and you can create any matrix you want but I’m not sure it really matters.

    I have my complaints with the CC but there are a lot of smart, experienced people using lots of polling information to arrive at these decisions. Just because they don’t publish their matrix doesn’t mean that they didn’t make one.

    The absence of evidence is not evidence of absence.

  2. Classic argumentum ad hominem, Dave. I offered both quantitative — albeit quick and dirty — and qualitative reasons to back up the claim that (a) Saanich is not the best choice, and (b) Guelph is a better choice. You offered nothing except to say, “I trust the campaign committee, ’cause they are… well… smart and experienced.” Uh huh. I’ll let readers decide which is more persuasive. Your defense of the CC’s wisdom couldn’t possibly be because your girlfriend is on it, could it? :-) The only problem is, your premise is wrong. The CC didn’t make this decision — it doesn’t tell Elizabeth May where she is allowed to run; Elizabeth made this decision herself.

    It doesn’t take a genius to figure out that starting from no IDed support and none of the riding canvassed to close a 21,000 vote gap in Saanich is more difficult than starting from with 5,200 IDed supporters and 1/4 of the riding canvassed to close a 6,500 vote gap in Guelph.

    How do you close the gap? By putting May in front of the cameras and hoping for the best? (and if so, would that not be just as effective in Guelph as in Saanich?) No, you need an effective ground operation, which Guelph has and Saanich doesn’t at the moment. Saanich could certainly get there if they put in the months and months of work that Guelph did, but you do realize that we’re potentially talking about an election only a few weeks away, right?

  3. Secondly, Dave, you mentioned riding polls. They are one of many useful pieces of data (and you will note that I arranged for these very polls before I left Ottawa), but they are insufficient as the sole arbiter of where May should run. They merely measure latent support, but don’t tell us whether there is any ability on the ground to convert that into actual votes (hence why the strength of the local team matters).

    And they are also somewhat flawed this far ahead of the election because we can’t even prompt for all the party candidates (since they haven’t even been nominated yet), and the results are only rough approximations because of high margins of error (a 400-person poll, as a cost of $7,000, gives us a range of 10% — i.e. a poll result of 25% means support could be anywhere from 20% to 30%).

    Oh, and let’s try to stick to the topic at hand here Dave, and not engage in false ad hominem attacks. Saying Greg Morrow has no campaign experience has about as much truth as saying Dave Bagler has a university degree. Both are, of course, factually incorrect.

  4. These are fair points, bluegreenblogger, although I think you give my quick and dirty analysis too much credit! One would hope that the CC, PCD, etc are doing a much more thorough job than that! But your point that such decisions should be data-driven is bang on.

  5. Hey Greg,
    Anybody actually picking relevant metrics, and ranking according to normative standards is worth taking notice of on Planet Green Party. Also, I wanted to get some more serious analysis of membership numbers vs. vote outcomes, but it wasn’t possible for me to get my mitts on clean current membership lists. My enthusiasm is partly envy over that.

    BTW, I do have a beef with the polls conducted. If, as you say, they were simple 400 respondent polls, then they’re not very valuable for decision making. Also, a simple ‘would you consider voting for…’ isn’t good enough either. Better to spend a whack of money on data that will really help plan the campaign, like responses of specific target demographics to policy stances, specific policy prescriptions, and campaign themes. That way, while it has cost a lot more money, you actually have a demographic roadmap, and strong guidance as to message, target audience, and ultimately the key to how to win, as well as where to do it.
    Good luck persuading anybody at the hub to pop six figures on the real deal though. Not when fuzzy warm feelings dictate strategy.

  6. “Oh, and let’s try to stick to the topic at hand here Dave, and not engage in false ad hominem attacks. Saying Greg Morrow has no campaign experience has about as much truth as saying Dave Bagler has a university degree. Both are, of course, factually incorrect.”

    Well, Greg, I hate to be rude but that was a pretty strong ad hominem attack on your part. Couldn’t you have skipped the last two sentences? The rest of your posts and comments have been more compelling.


    Anyway, I agree with you that data should drive the decision, but the final data crunchers in the CC (a) are not able to defend themselves or their position publicly with the internal data; and, (b) have not announced a decision officially. As an outsider to the final decision and having only seen so much data myself I cannot reach a conclusion on this matter.

    If I can also add one criticism of your numbers: Your ID index appears compelling, to say the least. Guelph at 100 and SGI at 2 is shocking. But, Guelph has IDed less than 25% of the riding, leaving 75% remaining. I might be inclined then to give Guelph an ID index value of, say, 25, to SGI’s less than 1, since any good ground campaign will still need to complete at least one full riding canvass. This doesn’t change the fact that Guelph still beats SGI, but it does make it closer. Given the narrrowing of the gap and the fact that you neither include every potential metric nor weigh them (they all have equal weight, which is arbitrary and possibly careless), I am not entirely convinced that your ranking system is complete and accurate.

    Yes, Greg, I agree: let’s have the numbers dictate strrategy – but only within reason. After all, you don’t compare that list with Central Nova. Given that riding’s members, votes, gap, and ID – all indexed and weighted in the way you’ve suggested – which riding comes out on top? So, I submit, you have chosen to eliminate that riding on the basis of factors outside of the factors you now propose we use to make our selection. I would thus propose that there are other factors we should use – both quantifiable and intangible – and I see that you agree.

  7. Full disclosure: I crunched numbers a few months ago to suggest a preliminary short-list. I used 12 factors, which were ranked and weighted. This created a short-list, not a final ranking.

  8. Mike
    Sure the riding matrix has it’s flaws. It doesn’t take into account everything that’s pertinent, but it’s directionally sound.
    You’ve really piqued my curiosity about what the CC is planning to do. You’re right, we really have no clue what they’re up to. Mikes right too, they cannot talk about what they’re doing, so it’s hardly fair to criticise them publicly based upon pure speculation. I should bear it in mind myself, because I have on occasion been a little callous.

  9. BGB — re: polling, I understood that EM needed convincing that CN was unwinnable but I felt they should only compare a couple alternatives. But EM wanted more options (“if I am going to uproot my entire life, I want to have more than one or two alternatives”) — including some highly questionable ones for flimsy reasons (“Glen Pearson told me he would campaign for me if I ran in London-Fanshawe, so I want it polled”; note: the GPC doesn’t even have an EDA in L-F).

    I thought these extra polls were a waste of money and instead wanted to leave some room in the budget for later polling to test campaign messaging, key policy positions, etc — exactly what you suggest. But that was poo-poo-ed.
    But hey, at least the polling was used this time, unlike before the CN decision, when over $100k was spent and the recommendations ignored :-)

  10. I’m going to weigh in with Greg on what I think is solidly based skepticism about what the CC, and/or May, might have put into the decision.

    The odds that they have some empirical imformation that would make up for that SGI/Guelph gap in ID’ed voters has got to be pretty low.

    The odds are far greater that there is no such hard information, and that everyone [or May pushing the rest] is relying on making up the difference with an excellent ground campaign.

    Dave’s comment that Guelph did not really ID that many voters cuts at east as much the other way.

    We have the people around May who seem to have done very little IDing in Central Nova, despite the tons or resources put in there over more than a year. No one has ever said different, and the indirect evidence points to that.

    So those folks don’t really have experience to build on, and even questionable what kind of campaign organizing potential there is in that pool.

    Then we have Guelph. Where it sounds like people went about it very systematically, as noted still didn’t get all that far even with substantial lead time. They could reasonably expect to go farther and get there faster the second time around… and possibly even somewhere else starting from scratch again. But to get that talent pool one or two whizzes moving to SGI isn’t enough.

    The likelihood of enough of those experienced folks moving to SGI seems really low under the best of circumstances, and with the antipathy between them and May and how she has dealt with them, seems remote.

    Good luck on that one.

    I will say that I can’t believe its very likely that even Elzabeth May is planning on a local air war doing the job. They had the local media eating out of her hand in Central Nova, and it did squat.

    Looks a whole lot more likely to me that May has really bought into the necessity of a top ground campaign, but in her quixotic way could not go where it would best for that [plus all the regional and national campaign factors in favour of Guelph]… and has convinced herself and enough of the rest that they are going to just ‘do it’ in SGI.

  11. And what is the price range of a house in Saanich these days? One that it wouldn’t be an embarrasment having her in.

    She’s unlikely to be holding and buying all these properties on her own dime. And virtually impossible she can qualify on her own for a mortgage, even if she could manange the cash flow to cover everything.

    If shes backed by family money she’s safe. But is there any or enough of that left. She has plenty of friends who could back her, but most if not all of them are also maximum contributors to her. And Elections Canada might take a dim view if one of them is financially backing a purchase [which is a loan].

  12. Yes my girlfriend is on the CC so I guess for that reason you could throw out my argument. Or perhaps since I don’t have a degree I guess I’m what? too stupid to figure out what riding she should run in?

    I didn’t say trust the CC. My point was that just because the CC didn’t publish a matrix or the numbers they used to come up with their decision doesn’t mean that they didn’t crunch numbers. The CC as it sits now was formed after the last election so holding them accountable for mistakes they had no part in making seems a little unfair.

    I really love how you asked me to stick to the point and called me out for using an ad hominem attack and then you follow it up by questioning my integrity and then an ad hominem attack of your own. Greg that’s a really nice glass house you’re throwing stones from.

  13. Quick thought on the GRIMES-gap between SGI and Guelph: Is it possible that SGI has its own database of IDed supporters that it chose not to enter into GRIMES?

  14. Mike, I’d say it’s almost a certainty that SGI has data squireled away. They have been doing semi-real campaigns for over a decade, and given the number of former Lib’s, and Dippers in the EDA, they must have known that it was important to ID the voters. I would suggest that most EDA’s that have been ID’ing the vote from 2005 election or prior will have a better laid out database app. than GRIMES, or civiCRM. There would be little utility for them to replace local d-base with a neat, but still somewhat deficient hosted app. Also, local jealousy, and particularism will mitigate against sharing any kind of data with the hub. I’ll confess here that I never shared all the ID’d voters in Etobicoke Lakeshore with the central Party. They’re all sitting on a hard drive in my basement, waiting to be put back into play.
    If the Party ever wants to be able to do direct mail en mass, or do any real analysis using ID’d voters, then the voter contact software should be beefed up to the point where it’s ease of use, but also versatility and user adaptability trump anything the local team might have in hand. Then people will use it, despite their jealousy, simply because it would be the best tool for the job at hand.

  15. And I can tell you from the experience of NDPVote, that versatility and broad use takes time and very sustained work at the centre.

    You don’t just plop out a database and expect it to be taken up.

  16. Even if SGI does have some significant voter ID, it can’t possibly be anything like what they have in Guelph.

    And again- what happened in Guelph was with definitely competent people who had some resources and time.

    Starting from scratch again, in a different place, with people who don’t have a track record and don’t inspire confidence…. is not exactly putting your best foot forward no matter how much money you are willing to throw at the problem.

    Just because May and the limited number of brains she trusts are [apparently] willing to acknowledge that the WAY they threw money at the problem in Central Nova was ill advised, does not mean they can get right what they have to do.

    That may sound like a truism, but its a common failing that people think that because they know what they did wrong last time, THIS time they are going to know what to do.

    Not to mention that IDing the requisite number of voters to have a good shot… would be a daunting task for anyone.

  17. Then, BGB, if that is the case, the ID-difference could be a lot closer. If we had that data available, the polling data to compare Guelph with SGI, polling data on second choices, and other information….. then I think it is conceivable that the data could be driving this decision, though Greg’s quotes above do certainly give me pause.

    But, let’s get back to the big picture. We need to avoid letting data entirely pre-determine future results. Look at Central Nova in 2006. The greens received 1.59% of the vote and were 23% behind 3rd place! In 2008, this was the strongest riding in the country. So, the introduction of Elizabeth May into a riding has an incredible impact on predictability.

    Further, let’s consider Gary Lunn’s 2008 victory versus 2006. A look at the numbers show’s that the Julian West saga significantly altered the outcome – mainly in favour of the candidate most likely to replce Gary Lunn. Now, we don’t know that all 2006-Lunn voters voted Lunn in 2008 or that the NDP moved to the Liberals or to the greens while 2006-Greens stayed home. We can’t know any of that. But, what we can say is that when a clear contender to replace Lunn emerges – even when only 3 weeks remain to coalesce around the challenger – that contender is able to get within 4%. AND, that contender was a small-g green. In SGI, it is conceivable that if there are a couple months of lead-time and a clear contender against Gary Lunn with strong environmental credentials, then that candidate has a strong shot.

    To sum up: (1) SGI might just be a riding where an enormous pro-environment, anti-Lunn constituency exists. (2) Introducing E May into a riding can increase the Green results by over 30%. (3) With twice the Green members as CN (and growing) and many experienced local campaigners, SGI might have the tools to exploit this potential.

    (1) and (2) are testable. Polling SGI to test for this eventuality – and comparing it to polls in ridings like Guelph – is the only way to check the hypothesis. I am going to suggest that data is the driver here and that polling data has shown that Elizabeth May can and will win in Saanich-Gulf Islands.

  18. The funniest thing about all of this is… just to give you a bit of history… when I was asked to join a CC call (then called the FCC) back in early 2007, I recommended that they look at Southern Vancouver Island in general and Saanich-Gulf Islands in particular from which to start building up a regional base, since that’s what you need to do to win in a first-past-the-post system (although GPC goesn’t quite understand that). Saanich and area has good long-term prospects for the GPC, but it takes time to build up a strong local organization, canvass the riding, ID supporters, etc. (which is why it is not the best choice for May right now, and why Guelph is a better choice).

    But Adriane Carr objected; she felt Saanich didn’t do very well considering all the money that was thrown at Saanich in 2004 (ironically, FAR more has been spent in Vancouver Centre — Adriane has a full-time salary and a full-time paid assistant, plus the BC organizer works out of her office — and yet she got about the same as Lewis did in Saanich in 2004). I suspect she objected mostly because she wanted to ensure Vancouver Centre was the go-to riding in B.C. Plus, she and Andrew Lewis had a history, so no doubt she did not want to see his riding be targeted.

    (Note from the GPBC’s Wikipedia entry: “Carr’s highly centralized leadership style had undermined efforts to broaden the party’s base of active members in the previous four years. Even after virtually all Parker supporters had left the organization, prominent Greens continued to depart citing an inability to work with Carr. The most striking example of this was Andrew Lewis, the candidate who received the largest number of votes in 2001…”)

    So Carr was dead-set against Saanich and said quite bluntly on that call, and I quote, “you don’t know what you’re talking about”. Funny, 2-1/2 years later, they’re finally heeding that advice. If they had actually done so 2-1/2 years ago, and worked since then to canvass the riding and ID support, then Saanich might well be the best choice right now.

  19. Mike – latent GPC support does not vary all that much across the country. Where we see differences, it is because work has already been done. In general, more so than the other parties, the GPC vote is almost entirely dependent on the resources (money, people) put into a given riding. The Greens targeted London North Centre and Central Nova, hence the exponential rise in support. But some of that rise is also due to the strength of Elizabeth May as a candidate — she can deliver personal votes — by my calculation, as much as an 8% bonus. But that’s true no matter where she runs. But think about what an extra 8% means in Guelph vs Saanich…

  20. There’s no doubt that SGI is a much better starting point than most other Conservative held ridings. It’s probably a nicer place to live than most other conservative held ridings. I can see all the personal reasons for Elizabeth to want to run there, and it is her life after all.
    You are right that the data can only carry you so far, and more polling can give a much better picture. If Elizabeth were running there for the Liberals, I’d be writing Lunn’s obituary right now. She’s not though, and a large proportion of Liberal votes will not be available to her. Sorry, but a great many people vote for the Party that they’ve always voted for. Just look at West’s vote, eventhough he pulled out, the dippers still pulled quite a few votes.
    Don’t forget to look at the turnout. The turnout was pretty big last time, so there’s not much room for growth amongst non-voters.
    The required vote needs to come from converting NDP, LIB, CPC, and Non-voters. It will be very very difficult, even impossible unless either a) the Conservative vote is suppressed, or b) The Conservative vote is won over. That means either a very positive, elderly conservative friendly message, or a strong negative message about Lunn. These things are possible, but are these facts recognised by the CC, and more specifically EMay? It could also be problematic with the local greens to campaign with a small c conservative message. We’ll see though, I’ve been wrong before.

  21. Greg,
    I’m pretty sure that the inner circle have been looking at internal leadership politics, and without doubt Adrian has had her eye on the leadership all along. I would judge that Central Nova, and Vancouver Centre were targeted as strongholds for future leadership, and internal political purposes. Your’, and anybody elses advice took second place to maintaining salaries, and control. As usual, the real reasons are concealed.
    I disagree somewhat with your low leadership premium for EMay. You have assumed that the money thrown into Central Nova, and London North Centre was effectively deployed. I was in London long enough to know that while they spent a lot of money, substantial amounts were simply thrown away. It was the hundreds of volunteers, Ben West’s activities on campus, the lesser importance of a by-election, and yes, a higher EMay premium that got the vote in LNC. From what I saw of CN campaign, they spent a whole bunch of money on local newspaper advertising, but actual voter contact got next to nothing. I think the EMay premium was quite a bit larger than 8% there as well.

    Once again, I would say the die is cast, and I’m turning my attention more to my local EDA, and next years leadership race. It would be very nice though to know what the national message is going to be though, because we need to be preparing our campaign, and determining our target polls for the general election here in Etobicoke Lakeshore.

  22. Greg: “In general, more so than the other parties, the GPC vote is almost entirely dependent on the resources (money, people) put into a given riding.”

    I agree to some extent on that and even produced a graph for this blog that looked at the Ontario riding spending and vote share. There was certainly a strong correlation. However, above a certain funding level, the few data points were all over the map. Are you certain that marginal resources continue to be as effective over, say, $40,000 as opposed to $5000?

  23. As to whether ever more resources deliver more:

    Obviously there is a diminishing return point. But depending on what you spend it on, not below the spending limit would be my observation, pretty much for all parties.

    Provided there is some minimum level of effectiveness to that. Spending more money on advertising and on a flagship presence probably produced no benefits in Central Nova after maybe 40-50K, let alone after over 200K.

  24. @BGB: absolutely, internal politics governed, I was just pointing out that I was asked to give advice, gave it, was told I didn’t know what I was talking about, then 2-1/2 years later, the very riding I suggested is now reportedly where the leader will run. It would be funny if it was so tragic that 2-1/2 years of on-the-ground work could have been done.

    re: May’s vote premium. You cannot compare a by-election with a general. I agree that the by-election premium is much higher, simply because more votes are personal ones and the government is not on the line. But in a general, it’s much lower. Put it this way: Ed Broadbent in 2004 delivered an 8 point premium in Ottawa Centre beyond what a purely statistical approach would project. I’m giving May the benefit of the doubt that she can deliver the same premium in Saanich as Broadbent did in OC.

    And you are absolutely right about party votes. In fact, typically 80-85% of people vote by party (about 2/3 by party, 15-20% by party leader/prospective prime minister). That will be lower in Saanich, in part, because the candidate matters more on Vancouver Island (see Keith Martin) and b/c a party leader is running (although May is not seen as local, so I’m not sure that local candidate mattering more actually helps or hurts), but still, you’re right that many Liberals will vote Liberal and ditto for the NDP.

    FYI, turnout was not especially high last time; in fact, 1,300 fewer people voted in Saanich than in 2006. I think a more important factor is actually the fact that Saanich is a large riding by population — 115,724. So there are more people in the riding; one could look at that positively, in that there are more potential converts, but I think it actually makes the ground campaign that much more important, and the Conservatives should have no trouble beating the Greens on the ground in Saanich, especially since they will probably receive more support from the central party this time.

  25. @Mike, Ken: I haven’t looked at 2008, but my 2006 analysis of spending vs results showed diminishing returns beyond $20k let alone $40-50k, although no doubt the leader’s riding is a special case. Every dollar spent yields less impact than the previous dollar. So spending $5k in 4 ridings would yield far greater impact than $20k in one riding, but of course, you need to go all out in ridings that you think you can win. But if the Greens were honest, there are only 3 campaigns that are even within 20% of winning: Saanich (but only if May runs there), Guelph (Nagy is not re-running) and Van Centre (Carr).

  26. That doesn’t disagree with my point Greg.

    Yes, there is a clear diminishing return in the GPC, and I’d agree it pappears to be as low as 20K.

    BUT, the bulk of those higher spenders spent the extra on advertising. In parties that spend more of it on the ground campaign I’ll bet the diminishing return point is at least twice that.

    And I think that includes the GPC- even though there isn’t much to go on since the larger spenders on ground campaigns can be counted on one hand…. and since you are so new at it, effectiveness of those ‘pioneers’ may be questionable.

    Making allowances for the latter, I see no reason why this will not in general be the same for the GPC…. not that I’m wishing you success, but thats another matter.

  27. Greg,
    I thought the turnout was up around 72%? I guess I was thinking about a different riding. Come to think of it, with the dipper candidate out of it, you’d expect the turnout to be exceptionally low. If the turnout was low, then it improves the odds for EMay. It is a big riding, both in terms of getting around on the ground, and sheer numbers. Both points mitigate against a late start for organising. I would hate to be the campaign manager. Imagine trying to build a dozen different strategies for every rinky dink corner of the riding, on the fly, without poll captains, or any prior fieldwork. Add to that a candidate who simply doesn’t understand that the candidates job is to follow direction, and Canvas, Canvas, Canvas.

  28. You don’t actually know for sure that they haven’t been doing fieldwork quietly in the riding.

  29. It may have been 72% but it was 73% in 2006. Basically, turnout was roughly the same. But Saanich is growing, so population is expanding. I disagree that a large population riding mitigates against a late start; quite the opposite!

    FYI, I did a post speculating on where things currently stand and a current best-case scenario.

    see http://democraticspace.com/blog/2009/08/can-elizabeth-may-defeat-gary-lunn/

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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: