5 replies to this topic

[Nick Springer]

Seen in Wal-Mart Era Wanes Amid Big Shifts in Retail in the Wall Street Journal

[frax]

It would be interesting to compare that to an index made out of population density and income. Likely WalMart locations follow that pattern, but where doesn't it, and why?

[Dennis McClendon]

Wal-Mart started in the mid-South providing big-city selection to small towns. But they embraced computerized inventory and distribution early, and proved better at it than the Kmarts, Ventures, Treasurys, Gibsons, and "catalog showrooms" already in the growing suburbs. Target had to go upscale in perception, and the others just went dark.

They are still very "underrepresented" in older, coastal cities. In fact, a good rule of thumb is that presence of Wal-Marts is in inverse proportion to whether the community had a fashion-oriented department store in 1975. If they did, very vew Wal-Marts will be found there even now. Places that didn't--small cities and fast-growing suburbs and exurbs--are where Wal-Mart has grown.

[natcase]

Thomas Holmes's paper, a cleaner version of the movie (as a WMV file) and notes are here.

He's here at the U of Minnesota (home turf of Target).

[Dennis McClendon]

So how would be a good way to map the relationship of Wal-Mart stores to population?

I suppose the obvious choice is to map stores per capita by state, with choropleth representation. That will show Arkansas and Oklahoma dark red and New York and Massachusetts light yellow.

But I feel that dot-density is a better way to represent this type of data, and I've been puzzling this through. Best I can come up with is to overlay store dots on different colored dots representing 1000 residents.

There must be some way to do it with a single variable, though. For instance, take population in each three-digit ZIP cluster and divide by number of stores in that same cluster, then plot one dot for every 1000 "shoppers"? Problem is that this will show where the Wal-Marts aren't, because each New York store will have 500,000 persons assigned to it.

What if we instead multiply stores by residents in each three-digit ZIP cluster, so the map shows "Wal-Mart choices available to residents."?

Other ideas welcome. I feel like too much statistical map work these days is lazy on the thinking about how to make the data meaningful.

[MapMedia]

I like the per-capita approach as I wouldn't want to know exactly where the store is, as with dots, but how many stores are available per person, and the story that tells:
Wal-mart friendly neighborhoods/zoning, competition, etc.

What interests me is the phenomena that paralleled and effected Wal-mart growth over the years (e.g., internet spending dollars vs big box dollars).