Sunday, January 27, 2013

Some research work I have been doing

For the last four months I have been working on a small research project.   I have been surveying people in BC about their voting behaviour.   I was not sure that I could get enough responses that I would be comfortable in writing up my research for public consumption, but I think I have enough responses now and sense there are strong enough trends within the data to tell us something.

I have asked 500 people in BC to take part in my survey and 364 have responded, that is much higher than I had hoped for.   My goal was to gain some better understanding of people's voting behaviours - how often do they vote, what are their motivations for voting or not voting.  

I have conducted a number of public opinion polls over my life and have training in the use and interpretation of statistics in the social sciences.   It means I understand the problems inherent in polling and what we can read from the data and what not.   I chose not to do a random sample opinion poll because I wanted to know the subjects well enough to have an idea of the veracity of their statements.   I did this because I believe that when questioned about the past many people do not tell the polling companies the truth.

In many opinion polls people are asked how they voted in past elections, if the poll is truly random their results should be a close reflection of the previous election turn out.  The reality is that this never seems to be the case.   What we get is something like 80-90% of the people answering the poll stated how they voted in past elections - that is much too high so either the sample is fundamentally flawed or the people are not telling the truth.

The reasons for the unusually high answer to how people voted in past elections is that people may honestly believe they voted (memory bias), they may confuse federal and provincial elections or they may not want to admit they did not vote (response bias).   It is for this reason that I wanted to have some connection to the people I talked to.

I decided to find people via social media to answer the survey.  I ended asking 250 people on Facebook to answer the survey, 200 from twitter or email and 50 in person.

Given the sort of people I know in social media, I needed to make a decision on how to factor the highly political.   What I did was that anyone I knew that was clearly active in BC politics in some form or had ever commented about my political postings would be excluded from the survey.  The 50 in person people were random people in Victoria over the last four months on the streets, buses or coffee shops.   They are the only people I had no previous connection with.  These decisions do not make the survey a random sample of the BC public.   I am entirely aware of this but I am only looking for confirmation work done by others and of anecdotal trends I am seeing.

My sample is geographically biased to the south island to parts of the southern interior - I tried to even this out but I knew from the start based on my contacts this would be unlikely.  

I tried hard to avoid too much confirmation bias but I know that there is likely some effect from this based on who I asked and how I asked.   I have inserted a selection bias into the work, but I do not think it caused a serious problem with the respondents, certainly self selection bias in online panel opinion polls is, I suspect, a much bigger effect and no one is calling any of the companies on it.

I am going to crunch the numbers and see what comes from it - I hope the data gives me something worth presenting.   There are a couple of things that did come to light and do sound obvious when I write them out:

People that vote in all elections are certain about having voted in past elections, but only about 1/4 voted in the last two provincial, three federal and two local elections.
People that did not vote all the time had a high degree of uncertainty about which elections they have voted in and this uncertainty seems to be higher further in the past.   There is clearly confusion among many people between provincial and federal elections in the past leading them being uncertain which one they voted in.
Only 43 respondents did not vote in any election at all - 15 of them were not Canadian citizens.
To the open ended question of "Why did you not vote?"  the main factors can be categorized as too busy, did not know enough to make a decision, did not like any of the choices.

Hopefully I will a chance on Thursday while hanging out and BC Children's Hospital to crunch the data and start writing up the results.   I will only write it for public consumption if I think the results can show clear results and are not random noise.
Post a Comment