A multiyear research project at Annenberg School for Communication, USC -- Metamorphosis -- attempts to identify Los Angelenos sense of belonging at the end of 21st Century.

The project has a communication focus; it is mainly concerned with how communication flows (media connectedness, transportation, telecommunication, computer networks) contribute to or deter the sense of community and belonging. One of the most exciting components of this project is mapping people’s sense of comfort with various areas of the city and correlating it with other spatial dimensions and with communication flows.

The mapping component of the project relies on hand-drawn maps and on geo-referenced information about people’s movements around town. Presently we have 200 hand drawn maps and information from about 2000 telephone interviews about people’s travel patterns in Los Angeles. The first stage of the mapping project, involving the hand drawn maps, involves digitizing the maps into a GIS format, which will allow us to measure spatial correlations between comfort/discomfort areas and various socio-demographic variables.

Metamorphosis is not just about maps. It is a richly textured socio-ecological investigation of Los Angeles, relying on a number of sources of information. One of the most important components is a random digit dialing telephone survey of respondents of specific ethnicity in several areas of Los Angeles. The survey is 40 minutes in length and it contains items related to media connectedness and use and community orientation. From the respondents who have completed the survey, between 10 to 20 were selected, for each area, to participate in focus group sessions. Besides being asked to map areas of comfort and discomfort, they were asked about their opinions on the cohesiveness of their community. A series of community leader and community organization activists interviews is also scheduled to be completed in the months to come.

The telephone survey mainly collects data about community cohesion and relationship with media. For example, respondents are asked about the factors that led them to move to the neighborhood; how often do they have discussions with their neighbors; what are the most important ways that they stay on top of what is happening in their community; what are the ways they get information to make decisions about what to buy.

They are also asked about their Internet connectedness and reasons for going on line. Media questions are especially interested in how people’s communication habits contribute to their sense of community. Thus, Metamorphosis wants to find how certain social goals are associated with various types of media, from the local to the national or even global. It is important to emphasize this component of the project because the analysis of the maps will be integrated into a web of media relationships, as one of the possible dependent variables of connectedness.

This presentation focuses on 58 hand-drawn maps collected during the first stage of the project. (The other 130 maps are in the process of analysis). Respondents used 4 colors to represent degree of comfort in Los Angeles. They were recruited from four communities: Crenshaw, Westside, South Pasadena and East Los Angeles (see map of location on next slide). The respondents were given about 20 minutes during focus group sessions to color in maps depicting in black and white the highway grid of the city. The maps also indicate names of major landmarks and neighborhoods. The maps were then redrawn manually in ArcView, by using the "create theme" function. Fifty-eight different vector layers were created. The process of transcription tried to replicate the exact contours drawn by the respondent.

Collected data: Westside, Crenshaw, East Los Angeles, South Pasadena

Partially Collected: Pico Union, Koreatown

Projected: Alhambra/Monterey Park

Respondents, both for the survey and for the focus groups, were selected from each area according to an ethnic screening procedure. In order to qualify for our survey, respondents in Westside and South Pasadena had to be racially white, those in Boyle Heights of Mexican origin, those in Pico Union of central American origin, those in Crenshaw of African-American origin, and those in Koreatown of Korean origin. The incidence rate varied between 70% and above in East Los Angeles, South Pasadena, Westside and Crenshaw and as low as 20% and 10% in Pico Union and Koreatown respectively. The survey has not been administered to respondents from Alhambra/Monterey yet because we have not conducted the interviews. The respondents in this area will be of Chinese origin, from the mainland, Singapore, Taiwan or Hong Kong.

When initially instructed to color the maps, the respondents were given the following directions: "Now I am going to give you a map of LA and some colored pencils. 1. I want you to mark your neighborhood in black first. 2. Then put green circles around the areas you know well and where you feel comfortable. Put yellow circles around places you do not know as well but go to at least occasionally and feel reasonably secure. Put red circles around areas where you feel uncomfortable, regardless of how well you know the area. Finally, put blue circles around areas which you just don't know. 3. Now color your area, using appropriate colors" (Focus Group Protocol).

Although the protocol refers to circles, in fact moderators varied in their instructions, most of them asking the

respondents to color in areas the felt comfortable or uncomfortable, according to the color scheme. Redrawing the maps in ArcView involved two different operations: creating shape files and adding fields to attribute tables.

ArcView vector maps are composed of a shapefile, georeferencing the the polygons representing a specific area and a database file (dbf), which associates to each polygon (feature) attribute data. In our case, each specific area on the map was assigned a comfort attribute value. The color scheme is translated into integer values. We have assumed that uncomfortable areas go in the opposite direction from those defined as comfortable. Thus, the negative value for uncomfort.

We are aware of the fact that the scale is not perfect, having a bias toward positive values. Ideally, the respondents should have given a five category color scheme, with a neutral point. Also, we should should have defined unknown areas as missing values.

After digitization, the maps look as represented in slide 8. In the process of redrawing we have tried to follow the street grid, using the highways and main roads as landmarks.

Once digitized, the maps are stored as shape files. In order to be able to collapse maps created by different people living into a specific area in order to create an average map, vector maps were transformed into grid maps. Using the Spatial Analyst extension of ArcView each individual vector map was used for creating a cell based map, as indicated in slide 9. The size and the number of the cells was specified in Analysis Properties, so that all the maps will have exactly the same grid structure. Each cell inherits the value assigned to the vector map polygon is located in.

The advantages of a gird map are immediately obvious. This is not just a map, it is its mathematical representation based on attribute values. Its most important advantage is that values assigned to the cells can be treated as any other mathematical values. They can be manipulated by using the four basic operations, loged, squared. In our case, the most important use of the grid map is its ability to be added up with other grid-maps and then divided by their number to obtain an average representation.

When graphically representing the average map thus obtained I used the z-score of each cell, instead of its raw score (see slide 11). This method is especially useful for comparing different average maps.

As mentioned above, average map colors do not represent raw scores. They are z-scores, representing variation of each cell, in either direction, from the mean.

Charts are useful for assessing normality of color distribution. Especially interesting are skewed distributions, which tell us if people in a certain area are on the average more likely to feel uncomfortable or comfortable in Los Angeles as a whole.

Three types of average maps were thus produced:

1. An average map for respondents from each area

2. Average maps for subsets of areas, based on people’s ethicity (e.g. White and Latino average map)

3. A "grand total" average map, based on all 58 maps digitized up to date.

The average individual area maps are useful for understanding how comfortable each group of respondents finds various areas of Los Angeles. Comparing the four maps thus created we can see where people of each ethnicity feel comfortable or not. The second type of maps are useful for understanding interethnic dynamics. The "grand total" map of Los Angeles is a work in progress. Its real value will be revealed when maps of respondents form all areas will be averaged in a final image of Los Angeles. This final map will then be used for measuring correlation with crime statistics, media exposure, and housing value in order to reveal what causes feelings of comfort/discomfort associated with various areas of the city.

The map above (slide 13) represents areas of comfort/discomfort for the 12 Westside residents included in the sample. This is one of the most polarized maps. The red values, of all the other individual group maps, are the furthest away from the mean. However, as indicated in the chart, the number of really extreme negative scores is low. The legend tells us that they do not go over 2 standard deviations. Yet, the dark red spot in the Compton area will have a great influence in shaping the "grand total" map, contributing decisively to making this the most uncomfortable area of Los Angeles. Interesting to note is the fact that the curve of score distribution, although skewed to the right, indicates a spike at the green extremity, strengthening the feeling of polarization.

Although South Pasadena residents live almost 20 miles away from Westsiders, there is significant overlap between the maps in slide 14 and the one in slide 13 (Westside). South Central, Compton/Wats are solidly red, another factor in scoring down the area in the grand total map. Also, there is an indication of positive orientation toward Long Beach area here, which is not present in the Westside map.The chart indicates that South Pasadena resident comfort scores are more normally distributed than those of Westsider. The curve is actually quite steep, indicating a tendency to converge toward the mean.

Combining the two white respondent area maps, by averaging them, we obtain in slide 15 a map with a lot more green,

although red areas are still solid and large. The map resembles pretty much both its source maps, showing the convergence of perception among Whites.

East Los Angeles respondents (of Mexican origin) seem to be the most isolated. The green areas in slide 16 are mostly oriented to the East side of the county and the shades of green in the other areas are rather faded and fragmented. However, it is worth mentioning that this could be also due to the fact that the respondents seemed less adept at reading maps, many of them resuming their mapping exercise to highlighting area names. Yet, very interestingly, their map converges with the ones for respondents in Westside and South Pasadena, indicating Compton and Watts as the most uncomfortable. Also, their map is skewed toward negative values to a higher degree than those of the White residents.

Crenshaw residents were selected from the Western part of what is currently called South-Central. They have covered (slide 17) most of the central and western side of the county in green, showing that they feel comfortable there. Compton and Watts are also considered as comfortable areas. East Los Angeles is considered, however, uncomfortable, as is Palos Verde. Although Pasadena is generally felt as being not comfortable, Altadena, a region with high Black population is indicated as a comfort zone. Glendale and the Valley also have considerable areas of comfort. Interestingly, the distribution of the scores is still skewed to the right, although the green-red ratio seems to be more balanced that it was the case for the East Los Angeles average map.

In slide 18 I have combined the average maps of both White areas with that of Crenshaw respondents. Remarkably, Compton, Watts, and a part of South Central appear now as uncomfortable. This means that the high score of the Crenshaw respondents were not enough to completely balance the negative scores of the White respondents. Also, the East-West divide becomes very clear. As I will show later, this map is almost identical with the "grand total" map of Los Angeles, averaged from all 4 area maps.

Combining the East LA with the White average maps in slide 19 preserves the same East/West divide encountered in the White and White-African American maps. In a test map, not shown here, a Latino and African-American map indicates that the two areas are complementary, but that they cannot agree on high disagreement areas, such as East-South Central areas.

Slide 20 presents the final, four map area of Los Angeles. It is easy to see that in its broad contours it resembles the map of the White participants. Also, the chart shows a pronounced right (negative values) skew. Based on this map we can say that people in Los Angeles feel less than comfortable in more than 50% of the areas of the city.

Summarizing the finding, we can draw the conclusion that, based on the present information, there is an obvious North-West / South East division in Los Angeles. The East-South-Central area of the city seems to be the the most uncomfortable and Westside the most comfortable. Significantly, and important for our future study agenda, Glendale seems to be an area appreciated by all respondents as comfortable, as it seems to be some areas of San Fernando Valley.

Many of the inferences presented above are quite subjective. They are educated guesses, following the direction in which the distribution of the colors seems to pan out on the map or in the charts. However, by employing the newly released S-Plus statistical package for ArcView one can measure the degree of association between various maps. The next natural step in this research will be to compare the maps by running correlations on their attribute table values. An alternative way to analyze the data is by entering the red/green area ratios for each individual map in the dataset resulted from the telephone interview. This would allow running more sophisticated analyses. They can reveal the causal paths leading to the degree of comfort or discomfort one feels in Los Angeles. A natural candidate for causal elements would be amount of media connectedness. This can be further refined breaking down the media into mainstream and local outlets, or print vs. tv vs. internet connectedness.

One important limitation of this study is the fact that the average maps are based on a rather small number of individual maps. In order to ensure statistical reliability we need to increase the N to about 30 maps per area. We intend to do this by sending maps to some of the respondents to our telephone survey asking them to color black and white maps, in a similar manner to the procedure used during the focus group sessions.

Map generation for this project will not be resumed to digitizing hand drawn maps. Based on a set of survey questions, which asked respondents to indicate the cross streets close the the areas they go most often or they try to avoid, we will be able to build point theme maps. These maps then can be used for further analysis, including spatial correlation and autocorrelation. Also, by using interpolation techniques, such as kriging, we will be able to generate continuous coverage maps. These maps, if converted to grid files, could then be used for comparing the maps generated from the hand drawn maps.

Another series of items in the survey asked the respondents to what degree do they visit each other’s neighborhoods. This data can then compared, at community level, to data obtained from the comfort maps. Statistical correlation could again reveal if the maps are associated or not.