Is the old adage about real estate also true for politics?  Recently the discount retail giant Costco Corporation sponsored an initiative to privatize liquor sales in Washington State.  I-1100 would have allowed stores like Costco to directly negotiate with liquor manufacturers for bulk purchases of liquor and then sell liquor directly to its customers.  [...]]]> Location, location, location,

Is the old adage about real estate also true for politics?  Recently the discount retail giant Costco Corporation sponsored an initiative to privatize liquor sales in Washington State.  I-1100 would have allowed stores like Costco to directly negotiate with liquor manufacturers for bulk purchases of liquor and then sell liquor directly to its customers.  Currently, all liquor sales in Washington State are handled through state liquor stores and the state negotiates contracts with distributors.  The state also collects a substantial mark up on the sale of liquor and taxes the sale as well.  A broad coalition opposed the measure which was accompanied on the ballot by a competing liquor privatization initiative I-1105 that differed from I-1100 in important ways.  The relatively broad “Protect our Communities” coalition ultimately outspent the Costco backed “Modernize Washington” coalition.    After all the votes were counted, Initiative 1100 was defeated by a margin of  46.55 % percent of voters supporting the measure against 53.45 percent opposing it.  This despite a late summer poll indicating that the measure was popular with Washington State citizens.

Why did a measure that once polled as popular ultimately fail?  Did the presence of two competing liquor initiatives confuse voters?  Were the active organizational efforts of labor unions (and other groups) effective in rallying “no” votes?    Was there a backlash against Costco’s public sponsorship of the measure?  Was it something else?

It’s always difficult in hindsight to reverse engineer a voter’s vote.  In direct citizen initiative politics many potential factors influence a person’s vote—possibly more so than in elections for office holders.  Initiative elections often involve complex and subtle variables including:  the initiative wording, the placement of an initiative on a ballot, misleading campaign messages by interest groups…etc.

One place we can look at in deciphering the results of I-1100’s defeat is where “yes” and “no” votes were cast.  Looking at a map of votes for and against I-1100 county by county in Washington State we see a clear regional pattern of “yes” votes that concentrates support in the most densely populated Western Washington counties.  “No” votes are clustered in the rural counties in the east, south and on the Olympic Peninsula.   This is a familiar voting pattern for Washington.  Washington politics often pit the interests of the populous and urban Puget Sound region against the more rural Inland Empire and Olympic Peninsula.   But a closer look at the counties that supported I-1100 reveals some surprises.

Washington State map with counties in green that voted for I-1100.

Two geographically small but rural counties (Island and Skagit) voted for the measure along with four sparsely populated, large and rural counties (Clallum, Mason, Chelan and Douglas).  Why did these rural counties support I-1100 while other rural counties voted against the measure?

During the campaign, some dubbed I-100 the “Costco initiative” because of  both Costco’s history in trying to deregulate liquor sales in the state and their public sponsoring of I-1100.  Costco invested heavily in the initiative campaign and was the biggest partner in a coalition that included other retail giants Safeway and Walmart.  Observers also noted that Costco promoted its support through their retail stores and memberships.  So is there a corollary effect of Costco locations on “yes” votes?  Let’s look at some data in support of this idea.

While we look at the data, lets focus on the six anomalous counties to see how the presence of a Costco in a regional shopping area might have benefited the initiative.  Before we look at rural counties lets take a at the three most densely populated counties first for some context.

We know that the largest concentration of Costco wholesale warehouses are located in King and Snohomish counties—these are the two most densely populated counties in the state that voted for the measure.  Pierce County voted against the measure but is more densely populated than Snohomish County.  Yet Pierce County has less Costco stores per person than north King county and Snohomish Counties combined.  The next biggest county by population to vote against the measure was Spokane County which does have two Costcos .  Although Spokane County does have one major population center, the county is not as densely populated as Snohomish, King or even Pierce (two Costcos) and Thurston (no Costco) counties.

View Larger Map

Of the six rural counties that voted for the measure two are somewhat densely populated—Kitsap and Island Counties.  Both of these counties are geographically unique because they are relatively small in area and have major Navy installations.  They also have access to at least one Costco Warehouse in the only significant shopping areas to service the counties.  There is a Costco in Kitsap County along the major Bremerton to Poulsbo retail corridor in Silverdale and there are several Costcos within 10 miles of the Clinton/ Mukilteo ferry run that services Island County’s South Whidbey Island area.  The northern Island County and San Juan Island areas of the county have at least two Costcos along the route to the Anacortes Ferry.

The four remaining rural and sparsely populated counties to vote for I-1100 give us a more interesting sample.  Clallum County on the northern edge of the Olympic Peninsula is the only county on the peninsula that has a Costco.  Bordering Kitsap County to the west is Mason County.  While Mason County does not have a Costco store, the only major retail-shopping district to service its population is the same Bremerton to Poulsbo district as Kitsap County–with a Costco warehouse in Silverdale.  A possible indication of the popularity of the Silverdale location is its components.  Not all Costcos warehouse stores are the same.  The bigger regionally important stores like their flagship store in Issaquah, WA. have gas stations, tire centers and food courts.  Not all Costco stores have these.  Silverdale does.

The two least populated counties to vote for I-1100 are Chelan and Douglas counties in the central part of the state.  Both of these counties are geographically large and very rural.  Their population centers both tend towards their borders with each other near the central Washington city of Wenatchee.  While Wenatchee is in Chelan County, it is just across the Columbia River from Douglas County.  This is also where one of only two major east west highways crosses the Columbia.  This is where the only Costco is between Spokane and the Puget Sound region.

Is there a correlation between Costco locations and yes on I-100 votes?  It is impossible to say for certain–remember Pierce and Spokane counties voted against the measure and they both have Costco stores.  Costco also doesn’t publish data on where their members live.

If we could map Costco members to location than we would have a clearer picture of how their campaigns might have worked geographically.  We do know that Costco targeted their members through direct mail and email.  We can reasonably assume that most Costco members live near enough to a store to justify the membership cost.  What if a Costco is the only big box discount choice in the region?  Would that justify having a Costco membership?  One assumption that we can’t afford to make is that all Costco members voted in support of the measure.

If we had access to other data in conjunction with our geographical data, we might be able to narrow our assumptions and come to a better sense of causality.  Data that might help us make a connection could include:

–the registration of light trucks, minivans and SUVs in an area–you need big cars to store large purchases from volume discount stores;
–the prevalence or convenience of existing state liquor stores; The membership in activist churches or civic groups opposed to I-1100;
–the prevalence and voting patterns of Native Americans whose communities have been devastated by alcohol.

The list is potentially endless.

One thing we can at least be curious about is why Kitsap, Island, Mason, and particularly Chelan and Douglas counties voted differently than their similar cousins.  Was Costco’s campaign effective in areas they had a direct influence over?  Possibly.  Maybe even likely.

Population density data provided by the Washington State Office of Financial Management:

http://www.ofm.wa.gov/pop/

We can thank Sir Humphrey Appleby of “Yes Prime Minister” for this humorous demonstration of how polling data can be manipulated to produce predetermined results.  Sir Humphrey’s fictitious example aside, political polls are an important part of both political campaigns and political journalism.  Because of their central role in American politics the need for solid data [...]]]>

We can thank Sir Humphrey Appleby of “Yes Prime Minister” for this humorous demonstration of how polling data can be manipulated to produce predetermined results.  Sir Humphrey’s fictitious example aside, political polls are an important part of both political campaigns and political journalism.  Because of their central role in American politics the need for solid data on the effect of polling data on the electorate is needed more than ever.

The volume of electronic media targeted at voters this mid term election is predicted to be the highest on record–significantly higher than the mid term elections of 2006 and even  higher then the presidential elections of 2008.  One estimate holds that 3 billion dollars could be spent on television and radio advertising nationwide this year.  While this is a boon for local broadcast stations, most viewers and listeners find the endless political ads annoying and alienating.  Despite their ubiquity and the amount of money that is thrown at their production, it is still (strangely enough) difficult to measure how effective electronic political ads are in garnering votes for causes or candidates.

I recently tried my hand at writing a voter survey (which you can see here).  While it was fun, it also showed me how difficult it is to create an informative, balanced and objective survey.  There are various reasons why survey writing is difficult (more difficult than any other writing I have done before).  To begin with, writing questions that are not leading is harder than it seems.  It is one thing to ask someone who or what they voted for, but to ask them why they voted that way in a simple question is not easy.  Do you ask them questions about what they believe?  How they  feel? or just limit your questions to what they do?  Just thinking of questions that are value neutral and informative takes a great deal of reflection.   Thinking through a question’s logic was enjoyable but also a little confusing–its a little like putting a jigsaw puzzle together without the guiding picture on the box.  You can always get it done, but you are never quite certain that you are getting the whole picture even though individual pieces seem to fit together.  They either fit together and are true to the picture, or they fit together but are not true to the picture–this despite the fact that you really want to solve the puzzle and see what the results are.

This is the promise of polling and surveys–the opportunity to look at something, measure the data and see something that is really happening.  But is Sir Humphrey correct that a “perfect balanced sample” is really just a fiction open to leading questions and manipulative context?  I’m not an expert on polling ethics, but I’d like to believe that with good research and proper safeguards, surveys and polls can be useful.

In developing my survey, I initially avoided the dozens of articles on survey writing you can find online with a simple Google search.  I wanted to really think about the questions I was writing and not rely on boilerplate formulas.  I wanted to see if I could develop reliable questions that would hopefully reveal why voters in Washington State voted a certain way on an initiative.  I narrowed my focus and looked at campaign sites for I-1100, I-1105, I-1098, I-1053 and I-1107 for background information and messaging styles.  I also looked at how much money was spent on various campaigns.  In looking at the various campaigns, I did notice a slight correspondence between money spent and votes garnered.  Some campaigns that spent more on advertising for or against an issue did relatively well at the polls.  The defeat I-1098 campaign spent over five million dollars and were rewarded on election day with a significant victory.   Some did not do well–the campaign in support of I-100 spent almost the same amount of money but lost.   While it is convenient to equate dollars spent with votes secured, it is not necessarily true–just look at voter reaction to expensive campaigns in the California gubernatorial and senate races.   The Washington state initiative races support the notion that we don’t necessarily know that much about the effects of media spending on voter choices.

In writing my questions, I couldn’t get very far before the doubts about questions neutrality surfaced.  Ultimately I was left with basic questions about party affiliation, actual votes cast, and basic beliefs about taxes and taxing authority–not very sexy.  But it was in the last questions about media influence that I started to see something of value.  These questions focused on the influence of media messages and venues and were based on a simple rating scale on how different media messages might have influenced a voter’s choices in all the initiatives.

How successful my survey is in collecting useful information about voter behavior in the 2010 election is probably negligible (particularly so because my sample is so small) but it does lead me to think about how surveys could be better.  Some questions I have after my experiment include:

–what kinds of questions work better than others in gauging voter’s attitudes towards issues?

–how can the effects of campaign messages be measured when it comes to voter behavior?

–how do you target a survey or poll to get the best sample?

–How many questions is appropriate in an online survey?

Some kind survey participants sent me encouragement and suggestions and I welcome those.  I  also encourage a broader discussion about style and methodology in polling below.  Please share your thoughts and suggestions.

A couple at the inauguration in Washington DC watching Obama take the Oath of Office.

It was nearly a year ago that Barack Obama was elected President. I remember election night as if it were yesterday and I hope I always will feel that way about that night. A year into his [...]]]>

A couple at the inauguration in Washington DC watching Obama take the Oath of Office.

It was nearly a year ago that Barack Obama was elected President. I remember election night as if it were yesterday and I hope I always will feel that way about that night. A year into his presidency and another year into the recession and another year fighting two wars, it is perhaps a good time to revisit the promise of last year and consider the continuing challenges ahead. Here are some more pictures from last years Inauguration in Washington D.C.