A Methodological Issue

For the practice part of the public policies, I am looking to see which urban designs contribute to social exclusion (see: Thesis page). This is usually called unpleasant design, hostile architecture, or defensive or disciplinary architecture. In the context that I am looking at, defensive architecture might be the best way to frame it; local governments implement defensive architecture to push out less desirable behaviours, like sleeping, skateboarding, or loitering. In regards to homelessness, any object that obstructs a person's actions that are usually directed at activities done by the homeless will be counted, so things like sloped benches, rocks under bridges, arm rests on benches, small openings on trash cans, bus shelters, stairs, etc.

There are no known studies that look at this other than ones that use propositional logic and qualitative methods. Qualitative studies are important in seeing if the individual believes they are obstructed by certain barriers, but it is important to quantify the amount in a numerical and spacial sense in order to see how excluding the local government is. In this sense, both are important to finding out which objects should be counted but also if there are enough of them around more desirable parts of the city. That means I must try to create a method for counting that is both valid and robust. It can also give future prospects for reliability, but that is beyond the scope of this study (the data, comparing four cities, an add to the reliability in a different paper).

I have looked at different methods for counting obstacles for people in wheelchairs, but it turns out most of them are qualitative and do not count them numerically. I have decided to use geographical methods and change them to suit a social science study. I need to count more than one area, within the city, between cities, and between countries; I also need to count more than one variable. The variables I should be counting are positive objects and negative objects. Positive variables are things such as awnings and long benches that lend itself to shelter and sleeping, respectively. A negative object in a bench with arm rests and small openings on trash cans that limit people sleeping on the benches and going through the trash to find bottles, respectively.

Here are the steps I came up with:

1. Find the average city block size and multiply it by two.

2. Grid the city in an uneven pattern.

3. Count each positive and negative objects in a stratified sample of grids based on more desirable areas, places were homeless people are seen, and places where they are not seen. Also, count the amount of visibly homeless that are seen separately.

4. Test to see if the areas are significantly different based on positive and negative objects found. If there is not, then there is no exclusion found. If there is, then move to step 5.

5. Then map where the homeless are usually seen (based on prior observation and interview answers). Be specific.

6. Transpose the second map onto the original map and count the grids that have any sort of homeless congregation in.

7. Finally, test the grids for any statistical differences.

Reasoning:

1. By having a larger grid to account for the synchronous blocks and control for any sampling bias.

2. By skewing the grid you can control for any sampling bias, for a similar reason in step one.

3. The stratified sample will be based on more desirable areas and places where homeless people are usually seen. Because the grids were placed in an uneven pattern, it is less likely they will be skewed towards any one particular sample; even the ones that have homelessness will be only a percentage and not the full square.

4. Two t-tests will be completed for each variable, one for positive and one for negative, between areas. If there is no difference, then that means the city is not actively trying to exclude the homeless from one part of the city. However, If there is an overall significant difference between positive and negative objects, then that means the city is excluding or not (depending on which is higher) for the whole city. This has other implications and can be measured with the other data gathered.

5. If specific places can be mapped, then it would provide a valid measurement to compare and to directly see if where they are seen are closer or farther from defensive architecture. The more specific one is, then there can be an increase in reliability.

6. Because it is known that homeless people do not congregate in all areas of the city, it is predicted that not all grids will contain homeless. A future study would focus on this issue itself and count the homeless over a week or so and then use the dot grid method of calculating the percentage areas. It is hoped this dichotomous result will garner valid results in order to compare.

It is my hope that this can be robust enough and provide some quantitative analysis towards how exclusive the community is towards homeless individuals. I will be doing some more research and testing of this method before putting it into practice, but if anyone reading this has any suggestions for reading, how to do this better, or any limitations, I would love to hear them. Please Contact me with your information.

#Research #Methods #Quantitative