Joh, Nguyen, and Boarnet on walking, built environment, and attitudes towards neighborhood safety

This manuscript has been out for awhile, but I’ve been too busy to read it with the care it deserved. Here’s the full cite:

Joh, K, M T Nguyen, and M G Boarnet. “Can Built and Social Environmental Factors Encourage Walking Among Individuals with Negative Walking Attitudes?” Journal of Planning Education and Research (2011)doi:10.1177/0739456X11427914

From the abstract:

We investigate whether the design of the built environment encourages walking above and beyond individuals’ attitudes toward walking. With data from a regional travel survey, we use regression analyses to examine differences in neighborhood walking trips among residents with positive and negative attitudes toward walking. The results show that built and social environment factors have a differential impact on walking trips depending on a person’s walking attitudes. Therefore, strategies to promote positive walking attitudes should be pursued in tandem with land use policies to encourage neighborhood walking.

They use data from a travel diary administered by the South Coast Association of Governments. The paper frustrates me somewhat, in that they leave out rather some (to me) important details. They stratify their sample into two groups: people who have a high preference for walking (the high walk group) and some who have a low preference for walking. They explain:

To compare these groups, the sample was stratified into two groups based on attitudinal differences, hereafter referred to as “high-walk” and “low-walk” attitudes. A walking attitude index was constructed based on an additive measure of three attitudinal questions from the South Bay travel survey. For each question, respondents were asked to rate on a 5-point ordinal scale (5 = very important; 4 = important; 3 = neutral; 2 = rather unimportant; 1 = not at all important). The median attitude index value of 10 was used as the threshold for stratifying the sample into high-walk and low-walk groups. Therefore, respondents who scored 10 or higher on the attitude index were included in the high-walk sample, while those who scored 9 or lower on the scale comprised the low-walk sample, with index scores ranging from 5 to 15.

Ok, but nowhere in the paper can I find what these three questions are, precisely. And that matters. Travel diaries tend to drive me crazy because transportation agencies seldom cognitively test their surveys with psychologists before administering them. I don’t know whether SCAG did so, but if they did, it would surprise me. Few people in transport are trained this way. This lack of testing isn’t something the authors of the study can do anything about; it’s not their survey. It’s just a problem that bugs me about agency-administered surveys in general.

But the questions they used to combine for their index should be explained in the article. Am I just missing it?

Splitting the sample means they are going to run into more trouble than otherwise when they begin to bring in their neighborhood level variables–which, for all practical purposes–work like another strata on the data. It’s not really clear what the split sample accomplishes for them that using the preference variables or indexes as controls wouldn’t accomplish just as well. The case for stratifying isn’t made clear to me anyway.

They are trying to figure out if built environment and safety variables produce more of an effect for those with different levels of preference towards walking in the first place. They are also, I suspect, trying to figure out if people sorted themselves by residential preference according to preference towards walking: those who very much prefer walking might be expected to, all things equal, buy houses in neighborhoods with more walking amenities. They don’t seem to see that effect here, which probably just means they can’t control for enough residential neighborhood variables to see an effect, or the effect is simply dominated by preferences for other things, like schools.

Their dependent variable is a count of walking trips.

These glitches notwithstanding, they have some interesting results, but what I find interesting, they don’t seem to find interesting. First, we’ll look at what they find interesting. They find that for those who have a high preference for walking, violent crime rates do not dampen the individuals’ tendency to walk much. And, in turn, a good supply of neighborhood walking destinations–a la the mixed use that planners promote so tirelessly–allows those with the preference towards walking to satisfy that preference. Yay! However, violent crimes seem to really reinforce lack of walking among those who show little preference for it–as we would expect–and no amount of neighborhood design seems to influence them to walk more. In fact, the Jane Jacobs idea–short blocks, lots of intersections thing seems to dissuade them! (I TOLD YOU PEOPLE SO! I’ve been saying it for years. Not everybody loves short blocks with lots of opportunities to get hit by a car!! Did you listen? NOOOO).

As long as I am behaving childishly, let’s look at the finding I think is interesting that they don’t comment on: the biggest factor that dampens the high walking group appears to be having children in the household. Don’t you think that’s interesting? I do. It’s a big effect size, too–bigger even than the effect that violent crime rates has on discouraging walking. The low walk group isn’t affected by having kids in any way. So if you want to stay with active transport, don’t have children. 🙂

It would be interesting to see–which they probably can’t do with these data–whether violent crime rates are informing the negative attitudes about walking to begin with.