Friday, February 9, 2007

Multifactor analysis of UBC VFM votes

I’m going on vacation to Mexico February 10-17, so here’s a going-away present for my readers: a toy you can play with while I’m neglecting this blog. It’s an Excel spreadsheet (fun, huh?) modelling how UBC media contestants won votes – linked here.

Some explanatory outline is in the spreadsheet. Here are a few more notes:

I created this to help understand what factors affected voting in the UBC media contest. I defined 4 factors:

Name = how well known the name is among voters
Serious = quality x quantity of serious election coverage
Fun = quality x quantity of fun election-related content
Non-web = amount of promotion by print media and advertising (including Facebook) beyond having a website

I subjectively evaluated each media contestant on each of these factors. The spreadsheet calculates an OLS regression to “explain” the nuber of votes based on these factors. I’ve posted the spreadsheet here so that you can download it, input your own subjective evaluations, and see what results you get.

Based on my numbers, the most important factor was “Name”. My preferred measure of importance is the number of votes (“Impact”) caused by a one-standard-deviation change in the factor. (Standard deviation here is the cross-sectional variation across contestants.) Impact of Name was 116 votes; of Serious was 50 votes; of Non-web was 38 votes; and of Fun was 10 votes.

Keep in mind that we can play around with the input numbers to make the outputs say anything we want – garbage in, garbage out – lies, damn lies, and statistics. I can’t claim any objective proof here. I’m just trying for some plausible explanation.

I think that a media group’s “Name” is affected cumulatively through time by its past performance, especially its past history of “Serious”, “Non-web” and “Fun”. So even though “Name” is the biggest factor in the short run, the other three are more effective in the longer run. (Of course there are bound to be additional factors I haven’t captured here.)

In my next post I plan to explore how media groups can become more successful.

1 comment:

Mike Thicke said...

Very interesting Mark. Makes me want to take another statistics course.

The only column that makes me a little queezy is the "non-web" column.