San Francisco 311 Data Visualizations: August 2013

Transcription

1 3 P a g e 1 San Francisco 311 Data Visualizations: August 2013 Saidah Leatutufu Mary Menees Kristen Wolslegel PA 755 October 9, P a g e

2 Introduction San Francisco has become an international center for technological innovation and startups such as Twitter, Dropbox, Eventbrite, and Square, among others. Karoly and Panis (2004) predicted that technology will accelerate in growth in the next 10 to 15 years, and that time has come (Karoly and Panis, 2004). Accompanying this growing tech-sector is the widely available amount of public data. San Francisco offers an open data portal ( to enhance open government, transparency, and accountability by improving access to data, with the stated goal of innovating how residents interact with government, resulting in social and economic benefits for the City, ( This portal is a major step toward e-engagement between the City and County of San Francisco and its workforce and residents. SF311.org meets the first and second objective of e-engagement through information dissemination as well as allowing constituents to manipulate data and provide feedback to the City and County (Reddick, p. 51). Currently San Francisco utilizes 311, a 24 hour customer service center for San Francisco residents. Constituents can reach a City and County service representative through the web, by telephone and by Twitter. 311 receives hundreds of reports each day requesting services regarding abandoned vehicles, illegal dumping, pot holes, graffiti, etc. The City then collects the data and posts it in an accessible format on data.sfgov.org. By providing public access to this data, the city wants to develop informed constituents moving toward the third objective of e-engagement: active participation in government. What does this 311 data tell the public? This data tells a story and analyzing this 2 D a t a V i s u a l i z a t i o n s

3 data makes room for a story to be told and solutions to be given for the enormous amount of problems this urban area is experiencing. For the purposes of this paper we have focused on one dataset, Case Data from San Francisco 311 from August 1, 2013 through August 30, We will present three visualizations that reflect the top 5 requests made to 311 according to source and location. The first visualization maps the GPS coordinates with the top five most common requests to determine regional differences in the types of problems reported. The second visualization displays the top five most common request types and the source that is used to make the request. The third visualization attempts to show a digital divide in the City and County of San Francisco. Upon analyzing these visualizations we will first, describe the methodology of how each of the visualizations were created; second, we will outline the major points gained from the visualizations; third, we will discuss the implications of this data and what it means to providing public service; finally, we will explain challenges and make recommendations of how this data can be used to improve the nature of San Francisco s public service. Data Visualization #1: Location of Service Requests The 311 dataset from August 2013 contained a very wide range of service categories. We selected the top five most represented categories, which comprised about 2/3 of the data from the month of August (10,135 of 16,667 total cases). These categories contained requests about illegal dumping, abandoned vehicles, sidewalk cleaning, graffiti (public & private), and damaged parking meters. 3 D a t a V i s u a l i z a t i o n s

4 Process of Creating Visualization: Various categories for similar complaints were merged so as to reduce the number of overall categories and to help observe general trends. For example, abandoned car categories for 2-door and 4-door were merged into one abandoned vehicle category and public and private graffiti were merged. Point coordinates were separated into longitude and latitude and data was imported into Tableau Software, a data visualization package that is user-friendly and accessible to the public. Position coordinates were overlaid with the five different categories on a map of San Francisco. The number of records was also included in the visualization, with a larger size of the point being proportional to a greater number of records. Major points gained from the data: There are a number of interesting trends that reflect different neighborhood characteristics. 1) Damaged parking meter complaints had a higher incidence in the Financial District and Civic Center than other city neighborhoods. These two regions are where most commuters from out of town work and are the most densely populated areas during business hours. It is therefore not surprising that these areas have the most parking meter complaints because they have the most parking meters. 2) Sidewalk cleaning complaints are most highly enriched in the Civic Center neighborhood as compared to all other San Francisco neighborhoods. This is possibly due to the high homeless population density contrasting with its broad visibility as the city s governmental center. Although there are homeless populations in other neighborhoods (Haight, Mission), these areas are not adjacent to City Hall, and therefore do not seem to be getting the attention that they may deserve. 3) Graffiti complaints are more prevalent in the Civic Center, Haight Ashbury, and Inner Mission. 4 D a t a V i s u a l i z a t i o n s

5 This speaks to the high population density and high traffic of these areas. Peripheral, lowerdensity neighborhoods are less plagued with graffiti problems. 4) Illegal dumping occurs evenly throughout the city with a higher concentration in the eastern part of the city, which has a greater population density. Trash pickup may not be frequent enough, or the high population density may make residents more careless. 5) Abandoned vehicle complaints are more diffuse than the other complaints, but there is enrichment in the Mission Bay & Bayshore areas which have a lower population density. This is possibly due to the fact that these are industrial, less developed neighborhoods where street cleaning and parking are less stringent. Implications and Recommendations: The type of 311 service request seems to have a direct correlation with the distinct neighborhood characteristics of San Francisco. The increased incidence of parking meter complaints in civic and financial areas where commuters enter the city to work should not be surprising. But whether the complaints are due to a greater proportion of damaged parking meters in these areas or simply because there are simply more of them cannot be resolved from this data. It is recommended that the city investigate the nature of these complaints and work to maintain proper parking meter function in these areas. The greater incidence of sidewalk cleaning in Civic Center tells a different story. Yes, there is a high homeless population there, but other neighborhoods have very visible homeless populations also. Why does Civic Center get all the attention? Probably because it is where the Mayor s office is located, as well as other high profile city offices. They need to present a clean and organized face to the world. It is recommended that the city be more consistent and fair in attending to the sidewalk cleaning of all neighborhoods, regardless of their status. In sum, 5 D a t a V i s u a l i z a t i o n s

6 visualization #1 shows that the unique characteristics of San Francisco neighborhoods determine the nature of 311 service calls. Data Visualization #2: Top Request Types by Source Process of Creating Visualization: The second visualization reflects the top five 311 request types by source (meaning, the method people used to contact 311). There are four methods that can be used to make a request: voice in or call, Open311 (forum), integrated agency (an internal office that makes the request), and online (Twitter and the web). Creating visualization that reflects the source that was used is important in order to determine how accessible 311 is and if the service is being used to its maximum potential. In order to create the charts according to source, the top five requests were collected from 311 case data including: graffiti, abandoned vehicles, damaged parking meters, sidewalk cleaning, and illegal dumping. Although these categories were widely represented in the dataset, in order to get a true understanding of how requests were being made, it was necessary to group relative categories. By using the top five requests, the visualization shows a clear picture of how the sources are being used among each category. In order to generate the data visualization, the case 311 data was imported into excel and was narrowed down to zip code, category (i.e. 311 external), type of request (i.e. graffiti), and source (i.e. voice in). Although the zip code was not necessary for this visualization, the numbers were needed for the visualization tool used which will be later discussed. After extracting the data needed from excel, it was input into the visualization tool, Many Eyes. Many Eyes is a tool that creates a number of visualizations like bar graphs and word clouds, and 6 D a t a V i s u a l i z a t i o n s

7 once published, this information is made available to the public. The case data was uploaded onto Many Eyes and a visualization was chosen. In order to really show the differences between the types of sources used, multicolored pie charts were chosen. Using the uploaded data, Many Eyes automatically generated the multicolor charts, the numbers of requests made and the percentage of the different sources used in each category. Major Points Gained from the Data: As represented in the visualization, the five pie charts represent the top five request types; each of the four colors represents each one of the sources. At first glance the most prevalent color is red (voice in). What does this mean? Obviously, most people are making requests via telephone, in fact, 6,007 people; but even more so people are not aware of or do not have access some of the other options. The request with the greatest number of voice in calls is damaged parking meters. Parking meters present one of the biggest issues in the city of San Francisco and there are thousands of them. As previously mentioned damaged parking meters are highly reported from the financial district and the Civic Center areas, 98.2%. Damaged parking meters can be reported for one of two reasons; one, a person reports the meter so they are not obligated to pay and can receive free parking; two, a person that has received a ticket calls in to report a damaged meter and contest the ticket. There is no surprise people are more likely to call in a damaged parking meter because there is a quicker response. Many people, when in a stressful situation, would rather speak to an actual person than to make a request on line, which explains why there were only 10 online requests. Perhaps, voicing in a request is the most used because there is a quicker response to the problem. 7 D a t a V i s u a l i z a t i o n s

8 Voicing in a request is the most used form than any of the other sources accept for graffiti. Interestingly enough, the top source used to report graffiti is an online method (33.2%). This could be for many reasons, but considering the amount of knowledge captured from each of the datasets it is likely that people are taking pictures of the graffiti and posting it online. 311 does not clearly define what graffiti is; it may be a very intricate mural or it could simply be a tag. The vague definition of graffiti could mean that 311 is misinterpreting graffiti posts. In addition to using voice in, integrated agency is the second most used source to make a request. An integrated agency is considered an office within San Francisco s public service sector such as the Department of Public Works. Integrated agency is widely reflected in reporting graffiti and illegal dumping. As previously mentioned, the Department of Public Works (an integrated agency) is responsible for maintaining the cleanliness of the city, meaning this department picks up any furniture left on the sidewalk, cleans graffiti off of bus stops and collects trash that does not fit into residential trash cans. It is expected that integrated agencies be associated with anything that has to do with cleaning up the city because it is their job, hence they are not called for abandoned vehicles and damaged parking meters. However, one might speculate why only 0.16% of sidewalk cleaning is reported by an integrated agency. It seems that most people voice in sidewalk cleaning so that compensates for the integrated agency requests that are not made. In summary, the request type by source visualization reveals that most people access the 311 service via telephone; damaged parking meters receives the most voice in requests and cleaning the city is managed by integrated agencies. 8 D a t a V i s u a l i z a t i o n s

9 Implications and Recommendations: It is important to consider how many people are aware of the 311 service and how many people are aware of the different sources that can make requests. It is clear that most people use telephones to make requests so what does this mean for the city? What if some people do not have access to a phone, but want to make a request? That is where the other sources come in. Since phones are widely used, 311 can focus less on advertising calling in and more on people Tweeting in or using mobile applications. Imagine how many more people could be reached if they could make a request from the computer at the library or their tablet. More advertising needs to be done for the other sources if 311 wants to expand its clientele. For example, when someone calls 311, they could tell the caller that tweet requests can be made if they do not want to be put on hold. Expanding the source list will allow more people to access the 311 service and in turn provide an easier customer experience. Visualization #3: The Digital Divide Process and Tools: The 311 data was downloaded from data.sf.org into Excel 2010 (for Windows) for review and manipulation. Second, the data was further refined by breaking the Point data into distinct longitude and latitude information for use in mapping the data. This was accomplished by inserting a column next to the Point column. Next the Point column was selected and using the Text to Columns tool under Data, separated the longitude from the latitude information into two columns. A few extra steps were required to remove unusable artifacts such as parentheses and commas, using the Find and Replace tool. Next, the same steps were used to separate the address column into Street Address, City, State, and Zip Code. 9 D a t a V i s u a l i z a t i o n s

10 Third, a filter was used for each column to allow selection of relevant data. This was accomplished by clicking on the row containing all the column heads and then clicking on the Sort and Filter tool. A drop down filter menu appeared at the head of each column. By clicking on the filter for Zip Code and deselecting null, extraneous data was removed. The data visualization tool used was Tableau. After creating an account on line and downloading Tableau Desktop 8.0, the Excel data was connected to a new Tableau work sheet by opening Tableau Desktop and selecting Data, then Find Data and selecting the Excel spreadsheet that had already been cleaned up and filtered. Once the data was connected, the Longitude data was dragged to the Columns and the Latitude data was dragged to the rows, then the Symbol Map was selected from the Show Me tool. To further refine the map, Zip Code was selected and dragged into Marks, along with Source. To populate the map, the Number of Records was dragged into Marks. To create the visualization, under Marks, Number of Records was selected using the Angle tool. The Number of Records was then selected under Marks and right clicked to pull up an editing menu, where Measure, Continuous, and Compute using Source were selected. Then Quick Table Calculation and Percent of Total were selected. Next Number of Records was again dragged into Marks and then clicked on the Size tool. Finally, under Map, Map Options, Data Layer, House Hold Income was selected by Zip Code using temperature colors. As a result of this data manipulation and visualization using Excel and Tableau, a map was generated that had multiple attributes that included: 1) a base map of San Francisco with areas colored from ochre to green showing 2009 Household Income (ochre equals highest income); 2) pie charts of varying sizes depending on the number of service requests per zip 10 D a t a V i s u a l i z a t i o n s

11 code (largest pie equals most requests); 3) pie chart also represents request by type (Interagency, Voice In, Web Self Service, and Twitter); 4) each pie chart is labeled with percentages that reflect the percent of web self service and or Twitter requests in comparison to all requests for that zip code. Major Points Gathered from data: Visualization three could be helpful to the public by indicating areas of the digital divide within San Francisco. The map suggests a relationship between household income and the digital divide. For example, in three zip codes, 94124, 94103, and the household income for 2009 was significantly less than the central and western areas of the city in zip codes 94110, 94118, and What the comparison shows is that in the lower income zip codes between 6.19% and 6.8% of all requests came by web selfservice, whereas in the higher income areas that percent was from 16.05% to 23.97% - a significant difference. This digital divide was not demonstrated in all areas and a map that showed more nuance in income would be very helpful. For example, in the area out west of SF State University only 2.61% came in via the website, but the sample size was three times smaller than the Bayview zip code, Visualization three could be more helpful than a data table to people reviewing the data, because it is easy to see at a glance the differences in income, the number of service requests, the source of those requests, and their relative percentages. A data table would be much more difficult for the public to see these underlying patterns. For example, the public could easily see that house hold income is lower along the eastern edge of the City compared to the central and western side. In addition, it is easy to see that the eastern side of the City has many more service requests than the western side. 11 D a t a V i s u a l i z a t i o n s

12 By enabling the public to better understand the data through this visualization, it could help to engender advocacy for free wireless service in San Francisco and or free text requests for service to 311 to reduce the barrier to online access for lower income areas. Implications and Challenges: Future data analysis could include more nuanced household income data, as well as the median age for each zip code to refine further the digital divide into age and or income. In addition, it would be very interesting to see if the number of Tweets grows. Currently the Mission zip code (94110) produces the most Tweets will that technology spread? Will a new communication technology enter into the scene? A challenge for the future is to advertise and include free texting to 311 as a way to submit requests because cell phones are nearly ubiquitous. A test pilot and tracking the data could show interesting results. If more people have access to 311 through a variety of communication methods, will it exceed the capacity of the City and County of San Francisco to meet the service level demand? How will the City manage public accountability? Conclusion The increase in freely available public datasets, combined with a rapidly advancing software industry developing open access tools has created the conditions for an explosion of visualization options for data miners in the public sector. In this report, we examined 311 service request data and found that there is a relationship between 1) neighborhood and type of request, 2) source and type of request, and 3) source and neighborhood. Interesting as these finding are, they reveal how an increase of information can have the unintended 12 D a t a V i s u a l i z a t i o n s

13 consequence of creating even more challenges than it solves. How is the city to reconcile the inconsistency in sidewalk cleaning or the digital divide in service calls? Integrating information into better public service is a challenge that can be mitigated through data visualization. 13 D a t a V i s u a l i z a t i o n s

14 Visualization #1: 14 D a t a V i s u a l i z a t i o n s

15 Visualization #2 311 Top 5 Request Type by Source Graffi Voice in 701 (24.8%) Online 937 (33.2%) Open (13.7%) Integrated Agency 791 (28.1%) Illegal Dumping Voice in 2,534 (64.1%) Online 260 (6.58%) Open (8.48%) Integrated Agency 821 (20.7%) Damaged Meter Voice in 880 (98.2%) Online 10 (1.11%) Open311 5 (0.55%) Abandoned Vehicle Voice in 825 (64.5%) Online 425 (33.2%) Open (2.26%) Sidewalk Cleaning Voice in 1,067 (89.8%) Online 70 (5.89%) Open (4.12%) Integrated Agency 2 (0.16%) 15 D a t a V i s u a l i z a t i o n s

16 Visualization #3 16 D a t a V i s u a l i z a t i o n s

17 References Karoly, L. and Panis, C., (2004), The Information Age and Beyond: The Reach of Technology. The 21 st Century at Work. pp Reddick, C. (2012). Public Administration and Information Technology. Burlington, MA: Jones & Bartlett Learning. 17 D a t a V i s u a l i z a t i o n s