NEW DEMOCRATIC PARTY OF CANADA
DOMITILLE BIEHLMANN. . . . . . . . . . . . . .260554000 JUSTIN CORBISHLEY. . . . . . . . . . . . . . .. . . 2 6 0 6 2 9 9 7 5 PIERRE-ANTOINE MOREL . . . . . . . . . . . 260564202 V I C T O R P O U R R AT . . . . . . . . . . . . . . . . . . . . . . . 2 6 0 6 9 9 5 5 7 KEVIN TRAN . . . . . . . . . . . . . .. . . . . . . . . . . . . ..260694403
TABLE OF CONTENTS
I N S Y 4 3 7 - M A N A G I N G D AT A & D AT A B A S E S PROFESSOR MAHMOOD SHAFEIE SARGAR PROFESSOR HANIEH MOSHKI WINTER 2016
OVERVIEW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 PROJECT OBJECTIVES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 CHALLENGES...........................................3 PROCESS.............................. . . . . . . . . . . . . . . . . . . . . . . . . . 4 ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 L I M I T AT I O N S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 CONCLUSIONS AND FINDINGS . . . . . . . . . . . . . . . . . . . . . . 9 APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
1
OVERVIEW BACKGROUND Before introducing the purpose of this project and the process that we went about to arrive to our conclusions, we begin with a brief introduction in Canadian politics. The politics of Canada function within a framework of parliamentary federal government. Canada’s electoral system is referred to as a “single-member plurality” system (“first-past-the-post” system). In all of the electoral districts, the candidate with the highest number of votes wins a seat in the House of Commons and then represents that electoral district as its member of Parliament. An absolute majority, that being more than 50 percent of the votes in the electoral district, is not required for a candidate to be elected. Any amount of candidates can run for election in an electoral district, but a candidate can run in only one riding, either independently or under a registered political party. In addition, each party can endorse only one candidate in an electoral district. Furthermore, Canada’s electoral system has since evolved in response to the country’s geography. Canada’s population, though it is not large in global terms, is spread over an extensive landmass. And as a result, some electoral districts are large though have lower populations. Additionally, The Riding Name Change Act, 2014 that came into force on June 19, 2014 modified the name of 31 electoral districts. For this project, our focus is on the New Democratic Party of Canada (NDP). We have been assigned to act as consultants for the NDP in response to their tough defeat in the most recent 2015 elections. PROBLEM After obtaining an impressive position in the 2011 elections, winning 103 seats (more than double their previous high) the NDP suffered a major loss in the 2015 elections. Being off to a great start in the first polls, they actually ended their campaign with 13% (44 districts) of the electoral districts in parliament, yet they won 19% of the total votes. This discrepancy between percentages has led the party to seek ways of establishing a long-term strategy. I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
Thomas Mulcair, the former head of the New Democratic party as of April 12, 2015, was convinced that he needed to put in place a long-term strategy. And since the NDP sustained a major blow in the recent elections, they are determined to make a comeback. NDP’s position, like in the last election, is to counter the Conservatives, without undermining Liberals. The 2015 election campaign and the impressive win by Justin Trudeau with the Liberal party began to raise the importance of data-driven decision making and fact-based strategy in election campaigns. Being the case, data teams have become valuable assets for election campaigns. The role for our project was to act as consultants for the NDP in order to gather insights from the data. Essentially, our team sought to answer the question of how to leverage big data and analytics in order to better position themselves in the coming elections? SCOPE In order to answer this question of how to utilize big data and analytics we needed to determine the scope of the project. In determining the scope of the project we need to address four key components: the 1) context 2) needs 3) vision 4) outcome. The context begins by defining the frame that is a part of the particular problems we are interested in solving: the overview and problem of this report cover most of the context for this project. The needs of this project are therefore to use data and analytics in order to provide the New Democratic Party with information that will assist them in their future campaigning. Furthermore, the vision for this project is to create a database to join the results from the past elections and demographics allowing us to analyze the vast amounts of available data. We will then establish relationships between social factors and elections results to determine which party an individual is likely to choose based on its social characteristics. The outcome for this project will then be to provide the NDP with an idea of what the main demographical characteristics of voters that they have the potential to captivate.
2
PROJECT OBJECTIVES
CHALLENGES
RESEARCH QUESTIONS
VOLUME OF THE DATA
Election laws are not reformed and as such ridings with a large proportion of swing voters are critical to campaigns. Once more, NDP’s implicit strategy will be to focus uniquely on those swing ridings that it has a chance to win in, while avoiding to weaken the Liberals in other swing ridings as mentioned before in the report. Therefore, our main research question will aim to answer who the main swing voters are and where they are located. More specifically we have outlined 3 main research questions.
The volume of the data was larger than we thought as we had demographic variables for each of the 338 districts across Canada and we had the voting results for each candidate of the 23 political parties for each district. Demographics Data included a lot of unnecessary information that we needed to remove as we judged they were not relevant for our analysis or too detailed. Once such example was the information of the languages spoken at home.
1. Where are the main swing ridings?
STRUCTURE OF THE PUBLIC DATASETS
The demographics dataset had years as 2. Who are the swing voters headers and variables as rows which was not com demographically speaking? patible with Valentina and Tableau. We needed to transpose the file while keeping the male/female Ie. What are the key breakdown by tripling the rows (See Appendix). demographically characteristics At the beginning, we wanted to incorporate defining swing voters? data from the past four elections (2006, 2008, 2011 and 2015) but the oldest two files did not have the 3. How comparable are the voters across same structure as the most recent ones some of the provinces? data was missing. The expenses file from 2011 had some missing data but the one from 2015 only included data for a few districts. Furthermore, the name of the We believe that there are many social factors districts in the 2011 expenses file did not corresuch as income, industry, education level, age and gen- spond with the names in the other files because of der that all play a role in identifying a swing voter, and the districts rearrangement in 2014. more importantly a swing voter that would vote for NDP. INTEGRATION ISSUES DATA SOURCES For our data sources we used a large variety of public data, including data from the government of Canada through Statistics Canada, Election Canada, as well as different census data. Firstly, we obtained the Official Elections results of 2011 and the Official Elections Results of 2015 from Elections Canada. The legitimacy of this source ensured reliable and accurate information to provide the base for our project. We then retrieved our Demographics Data for 2013 from Statistics Canada, an equally reliable and accurate source. I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
Our different datasets were supposed to be joined by district number when it was possible, but mostly by district name. However they were not consistent from a file to another (e.g. GlengarryPrescott-Russell in 2011 became Glengarry-Russell-Prescott in 2015), as well as with French names (QuĂŠbec/Quebec), creating difficulties to link our tables and needing manual intervention for every discrepancy. This issue also occurred for the expenses files and could not be linked with neither the demographics nor the results data
3
PROCESS GATHERING THE DATA SETS As mentioned in the data sources section, we gathered data from Elections Canada and Statistics Canada. We retrieved files for the 2011 Election Results and 2015 Elections Results as well as the Demographics Data from 2013. These three files were all in CSV format. Considering that there are 338 electoral districts - these files were very large for both the electoral results and demographics. Notably, the demographics file contained many records including information on work hours, education levels, and other specific information rendering it even larger in size. We initially started off by creating a miniaturized version of our database by including exclusively Mont-Royal and Montcalm as our main ridings. This allowed us to test how the importation for the next 336 ridings would happen, and what particularities would arise. We therefore imported these datasets, encountering multiple issues along the way which will be discussed throughout this paper. We therefore initially imported the data on the 2011 electoral results based on the 2013 geographical distribution of ridings. This was to allow us to compare these ridings with the 2015 results, which saw a change in geographical distribution from 2011. We then imported the results from the 2015 elections for these two ridings, followed by the importation of the demographic data for these two ridings. We chose certain key demographic aspects of the population we believed would be more relevant. This allowed us to eliminate some of the less relevant demographic information we had. We decided to keep the repartition of the population in terms of their jobs, their age, income disparity, unemployment rate, as well as their education level. Once we finalized our reduced version our dataset, we proceeded to import every single riding in Canada. The importation was based on how we imported the initial two ridings (Mont-Royal and Montcalm). After gathering the necessary datasets, we needed to extract, import and clean all of the data. In order to do this, we used a variety of different softwares including Excel and OpenRefine. Additionally, we utilized some VBA coding within the process. The fact we were working with public data made the manipulation of the data slightly more complicated, since this is generally not data that is well built or arranged. In order to create I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
a manageable database, the demographics data needed to be transformed extensively. We needed to transform the dataset because we wanted to move the demographical characteristics from rows into columns. Moreover, the distinguishment between male and females was not manageable with the CSV file. We need to triple the columns for all of the demographical characteristics that we used for the study. CREATION OF THE DATABASE After a lengthy process of extraction, importation and cleaning of the data sets, we were then able to create the database. We created the database on Valentina allowing us to structure a relational database between the various files. We used SQL in order to link our tables together through the unique federal number of each riding. This allowed us to have a well structured database. We also imported data on the expenses for the different parties, divided by region, but then realized it would not be possible for us to link these tables to our current database, since the expenses were not applied to the same geographical distribution. ANALYSIS We began the analysis process following the completion of the database. Once finalized, we were able to upload the database into both Tableau and CartoDB. This allowed for us to answer our initial research questions such as identifying swing ridings for example, or identifying key characteristics of swing ridings. We were able to confirm or reject our initial hypothesis we had made at the very beginning of the study. We were able to analyze geographical data with CartoDB, while analyzing proportions and distribution of the population in Tableau. We decided to chose the ridings in which the NDP lost the most votes from the 2011 elections to the 2015 elections, and compare the characteristics of these ridings to the rest of the ridings in Canada. By operating in such a way, we were able to determine who the mass of voters who decided to change their votes were and what are their characteristics. We were able to achieve conclusive evidence in our analysis, allowing us to make interesting conclusions which will be further discussed in our analysis section of the report.
4
ANALYSIS In this part of the report, we will try to provide elements of answers to our main research questions. After a careful review of our result, we decided to focus our attention to key strategic areas, rather than providing results to the entire country. By looking at the location of swing districts, we came to the conclusion that most of the swing states were located in Quebec, Ontario as well as in other parts of New Brunswick.
ing previously won districts. But the most significant blow comes from the NDP, who lost more than half of its districts. One key insight here, is that Liberals reappeared as a strong force in the political landscape by essentially winning back NDP districts. As a matter of fact, the NDP lost 73 of the districts won in 2011, 59 of them going to the Liberals. Interestingly, the 14 districts that did not go from NDP to the Liberals are all located in Quebec, clearly giving evidence of the strategic aspect of the province. Additionally, we can OVERVIEW distinguish an increased presence of an outsider party: Bloc Québcois. Indeed, the party has managed to gain more dis On Cartodb, we used an SQL function to display tricts in Quebec. It is important to mention that, visually election results in 2011 and 2015. We added a layer to dis- speaking, we must not be biased by the large amount of land play the color of the winning party: blue for Conservatives, covered by Nunavut and Yukon, as they have a significantly red for Liberals, Yellow/Orange for NDP and Green of the smaller amount of districts and population. Green Party and purple for Bloc Québecois. We believe this visualization gives a good overall idea of the results, and Consequently, the Liberals clearly reversed the balhelps us understand each party’s importance in the Cana- ance of power by winning roughly 40% of the votes, comdian political landscape. pared to the 32% won by Conservatives. Similarly, the NDP lost a third of its presence in Parliament by now becoming MAP 1: 2011 RESULTS (APPENDIX 1) the third most represented party in Canada (20%), a position previously owned by the Liberals. Visually, we distinguish a predominance of blue, meaning that that the Conservative party won a large chunk MAP 3: SWING DISTRICTS of the districts in 2011. By the numbers, this conclusion is confirmed since the Conservative won the 2011 election We entered a specific query within Cartodb so as to by obtaining roughly 40% of the votes. Another key insight highlight the districts that changed between 2011 and 2015. from this map is the result of the New Democratic Party This allows us to better identify the key regions in which (NDP, yellow on the map). Indeed, the NDP seriously chal- most swing districts are located. We realize that most of the lenged the conservatives, by winning about 31% of the votes. changes occurred on the eastern part of the country, particThe left-wing party was comfortably seated as the second ularly in Quebec and Ontario. More than half of the districts party of the country and leader of the opposition. The same changed between the two elections, clearly highlighting the cannot be said on the Liberal, who suffered from a signifi- inconsistency and fragmented dimension of the votes in cant blow by regrouping only 19% of Canadians. Canadian elections. Only 166 districts kept their political The political landscape was clearly dominated by affiliation. The NDP accounts for a good number of those the conservatives in 2011, despite a strong presence of the changes by failing to convince the ones that trusted them NDP, who clearly positioned themselves as the main oppo- in 2011. We could conclude that, by solely looking at NDP, sition to the right-wing party, poorly backed by the Liberals. the representatives elected did not offer enough solutions to Canadian citizens that voted for them. MAP 2: 2015 RESULTS (APPENDIX 2) If we carefully look at Quebec and its main city Montreal, we see a large amount of swing districts within By simply looking at the map and comparing it to the greater Montreal area but not at the center (Appendix 3). the 2011 election map, we can clearly see the bounce back This multitude of swing states in the area must be addressed from Liberals. Clearly, the conservatives had issues keep- by the NDP because they represent districts in which the
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
5
party lost more than half of their votes between the two elections. The party must identify the reasons why the voters that chose them in 2011 decided to trust the Liberals in 2015. Before getting into the elaboration of the appropriate political message, the party must identify those voters in term of demographics. This is what we will do in our different analysis below.
political message to the “native” population, very present in those states. We can also argue that more strategic provinces are more sensible to other variables (social policies, immigration policies etc.) than less populated area whose main concern are job creation. ANALYSIS #2: MANUFACTURING SECTOR
ANALYSIS #1: DEMOGRAPHICS (UNEM- As mentioned above, we decided to narrow the foPLOYMENT RATE) - APPENDIX 4 cus of our analysis to identify strategic areas. Here, our focus will be on the greater Montreal area. Indeed, Canada is not Our first intuition regarding correlation of variables homogeneous enough in terms of swing ridings. That is why and votes was to use unemployment rate as a variable. We we decided to solely focus on the strategic province of QC, thought that unemployment rate would be a good indica- in which many districts changed their political color. For tor of the overall political state of the districts. For example, this analysis, we have identified the following districts: we believed that a high unemployment rate would lead to a Mirabel, Terrebone, Montcalm, Rivieres-des-milles-iles, change in political orientation of a district, as people would Therese de Blainville, Chateauguay-la-colle, Saint Jean not be satisfied by their representatives’. On the contrary, a The total population of those districts is approxilow unemployment rate would yield satisfaction and thus mately 711,095, thus a significant amount of canadian votpolitical stability in the district. ers. Additionally, we chose these districts based on their Our analysis helped us realize that part of these as- 2015 voting results. Indeed, the Tableau dashboard we have sumptions were right, but did not constitute a viable answer created helps us understand that the NDP lost about half of to NDP’s problem. What we observed, is that high unem- their votes in the selected districts. We will try to give an anployment had a good correlation with swing districts in par- alytical and strategic recommendation to the NDP in order ticular provinces. As such, New Brunswick has many swing to help the party win back those disappointed voters. Using districts and the map shows a lot of red in this province, Cartodb and Tableau, here are our analytical results: meaning that the unemployment rate was high in this area. The same applies to Nova Scotia. Those regions have similar CARTODB (APPENDIX 5) characteristics regarding native population. Indeed, those regions have a high proportion of amerindians. This partic- The Montreal downtown districts did not change ular segment might be more responsive to high unemploy- and have a low presence of manufacturing workers. Comment rate and we recommend that the NDP addresses their paratively, as we gradually move away from the city center, issues in their political program for the mentioned regions. suburbs districts changed their votes and have a higher conThere are many of them, meaning that the NDP could cap- centration of manufacturing workers (20% of all employing ture a large amount of undecided voters to join their ranks. industries). Since those suburbs changed their votes and Unfortunately, the analysis made on unemploy- that there is a larger proportion of manufacturing employment is specific to New Brunswick and Nova Scotia and we ees compared to other employment types, we can conclude were not able to apply it to the entire country. As a counter that employee of the manufacturing industry are key voters example we could take the case of Toronto, Montreal and for political parties. Vancouver, three of the largest Canadian cities, where the Additionally, we can make the assumption that unemployment was somewhat low, but also where districts manufacturing workers are more highly concentrated in the swinged. suburbs because of their lower income. By conducting an Consequently, we cannot argue that unemploy- analysis with the 30,000-39,999 income range, we notice that ment rate is a viable variable to explain why voters ended up the number of people with this range of income represent up changing their political orientation. But we can definitely ar- to 30% of the population in some districts. If we added the gue that if NDP wants to win key swing states (New Bruns- 20,000 to 29,999 income, this number would reach 50% of wick, Nova Scotia in this case) it will need to address their the population (Appendix 6). Based on this conjoint analyI N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
6
sis, we can come up with the conclusion that manufacturing workers are key swing voters. Indeed, the districts in which they represent a majority have switched between elections. We added the income variable to confirm our assumption that they were not present in the downtown districts where average income is somewhat higher.
ANALYSIS #3: FOCUSING ON DISTRICTS WERE NDP LOST MORE THAN HALF OF THEIR VOTES
To start this analysis, we picked the districts in which the NDP lost at least half of their voters between 2011 and 2015. The goal here was to develop a demographical profile TABLEAU (APPENDIX 7 & 8): of the voters that changed their minds regarding NDP. Acadie-Bathurst, Gatineau, Hull-Aylmer, Louis-Saint We have created two dashboards on Tableau. One Laurent, Marc-Aurele-Fortin, Repentigny, Saint-Mauricegives us the cumulative demographic data on the selected Champlain, Shefford districts and one with the demographics of the rest of the canadian districts. We used those two tables in order to com- TABLEAU: APPENDIX 9 & 10 pare the demographic profile created for the selected districts with the overall profile relative to other districts in Canada. With the help of two dashboards, we analyzed both This allowed us to make a pattern stand out and not confuse the cumulative data from the above-mentioned districts and it with the average profile of the Canadian voter. Thanks to the overall demographics of Canada. the dashboards we have created on Tableau, we were better On the cumulative dashboard, we observed that able to analyze the different demographical characteristics most of the voters in these districts worked in the public of the districts we have selected. This dashboard displays the administration. To contrast this result, we looked at demoinformation on income ranges, educational level, age ranges graphics from all the other districts. The reason why we and leading employment categories. Additionally, we have made that comparison is because we wanted to highlight decided to display the results of various political parties in particular differences that stand out between the set of disorder to highlight gains or losses in votes from 2011 to 2015. tricts we have analyzed and the rest of the canadian districts. The NDP lost more than a third of their votes in this area Thanks to this, we realized that the public administration (-35%), where manufacturing industry is the most rep- sector was very present in the districts in which NDP lost resented employment category. Additionally, 25% of the more than half of their voters between 2011 and 2015, much population do not have certificates in those districts and the more than in the rest of Canada. We could make the concluaverage income tends to be located in the lower-middle class sion that, for example, workers from the public sector have segment ($20,000 to $29,999). But comparing to Canada, been disappointed by the NDP’s actions for their interests in we observe that the selected districts have average incomes those districts and chose to give their votes to the Liberals. below average, with the manufacturing industry representa- The key takeaway of this analysis would be to say tion being below national average. Under the same idea, the that if the NDP wants to win back the employees from the average number of citizens without certifications reached public administration, it should adapt its political program 20%, compared to 25% for our selected districts. Conse- to seduce the public administration employees quently, education level is lower on our selected set of districts, as well as the average income level. We are now able AFTERWORD to confirm the hypothesis stated in the Cartodb analysis and also develop a profile swing voter for these regions. We could do many more analysis regarding demo graphic profiles of swing riders. In fact, thanks to our tools, The takeaway here is that the NDP should target the NDP could analyze each swing district and determine a employees from the manufacturing industry in the suburbs profile that it should target (rich/poor, high education/limof Montreal. We have proved their presence in the districts ited ecution, agriculture/manufacturing‌). We have highthrough income, educational and employment sector analy- lighted the few that we believed were most important in the sis. By benchmarking our results with the canadian average case of the NDP. profile, we were cable to confirm this profile has being exclusive to the districts selected, and thus relevant to the NDP. I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
7
LIMITATIONS FROM THE PROJECT PROPOSAL INITIAL INTENTIONS
EXPENSES TABLE
In our initial project proposal, we were extremely ambitious in what we wanted to achieve with the amount of time available to us. We initially wanted to use PowerMap to do our visualizations of the different election results, but we ended up using CartoDB instead. This allowed us to have the entirety of our data in the data for the map, while allowing us to manipulate and visualize the data with far more ease. We had also mentioned we would compare the election results from 2006, 2008, 2011 and 2015, but we quickly realized this would be far too ambitious, since the election results from each year were not always structured in the same way, while also having a different distribution of ridings. Therefore, we only used the election results from 2011 and 2015 to come to conduct our analysis.
We had also mentioned in the proposal we would create an expenses table that would sum up the individual party expenses by riding. Even though this seemed like a plausible idea and initiative at the time, we realized throughout our study that the structure of these files made it extremely difficult to use. Many of the ridings were not identified properly, while many ridings simply had no data at all. This made it extremely difficult to use, since we couldn’t link this table with our other tables properly due to the important lack of accuracy of the data. However, we did manage to use the other dataset we had mentioned we would use, the unemployment rate. We were able to get the data about the unemployment rate in canada by riding, and were able to link it to the rest of our dataset. This led us to conduct an in depth analysis about this economic factor, and reached conclusions that will further discussed in this report.
URNS Also, we had initially thought of using the information from the urns in order to be as specific as possible in our analysis, but this turned out to be nearly impossible, since there were no distinguishing factors between two urns from the same riding. This made the use of urns useless to us, since the only relevant information we could use from this dataset would be the aggregate data of each urn, data that was already there in the dataset from the elections. LINKING We then proceeded to import the demographic data from 2013 just as we had planned in the proposal, but the linking this dataset to our data on elections was not exactly as we had planned. We initially thought we would be able to link both these tables through their district name, but as it turned out, some of the names were not always properly written between tables, which forced us to come up with an alternative. We realized that the federal numbers of each district were present in each table, which allowed us to link them using this attribute. I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
RESEARCH QUESTIONS In terms of our research questions, we managed to meet all of them except for one: “Which ridings receive too much funding and which receive too little?” The remainder of our research questions, such as “What are the demographic characteristics of swing voteres?”, or “Which ridings can be considered swing?” were fully answered in depth with a strong analysis to support our conclusions. TIMELINE Finally, we fell slightly behind on our timeline. Our month of February went as planned, but the month of March did not go exactly as planned. We had initially underestimated the time we would require in order to have a fully functional database we could use to conduct analysis on the elections. This process took us far more time than initially planned for, which pushed our entire study to be longer than anticipated. Rather than conducting our analysis in March, we conducted it mostly in April.
8
CONCLUSION In conclusion, our project turned out to be quite successful. We found interesting conclusions answering our different research questions, including “Which ridings can be considered swing?”, “Who are the swing voters from a demographic standpoint?”, and “What sort of profile do these voters and areas have?”. We have conducted in depth analysis in order to answer these questions, and we have found significant evidence that supports our conclusions. We are able to confirm that, unlike public opinion, the unemployment rate in ridings does not affect voter trends, since there is no significant trend with regards to this characteristic in ridings. Also, we have determined that the middle class employees working in the manufacturing industry have a significant impact on voting results for the NDP, since we have determined that this key demographic is an important aspect of the voting population that pushed the NDP to lose such a significant amount of ridings. Finally, we have also determined that the voters who work in the public administration sector are very important voters for the NDP to target. By adopting policies and modifying their speech to cater to this demographic, the NDP will be able to gain a significant amount of votes at the next elections, and will be able to bounce back as one of the leading parties in government.
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
9
APPENDICES
Appendix 1: 2011 elections results by districts and political affiliation
Appendix 2: 2015 elections results by districts and political affiliation
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
10
Appendix 3: Swing ridings between 2011 and 2015 (Montreal)
Appendix 4: Unemployment rate level by district (more red= higher rate)
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
11
Appendix 5: Manufacturing sector in Montreal (more red, more employees from the sector)
Appendix 6: Income $29,999-$40,000 same region
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
12
Appendix 7: Selected district demographics We decided to use a pie chart to show the reader the proportions of each employment industry For industry: Red: Retail Trade Dark Green: Manufacturing Light Green: Wholesale Trade Light Orange: Construction Light Grey: Health Care Dark Grey: Educational Sky Blue: Public Administration
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
13
Appendix 8: Demographics for entire Canada
Appendix 9: Cumulated demographics on districts where NDP lost more than half of their votes
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
14
Appendix 10: Cumulated demographics on rest of Canadian districts
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
15
A screenshot of the ERD of your final data structure, and a data sample
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
16
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
17
LINKS CARTODB LINKS Swing_ridings+demographic (unemployement) https://benjovy.cartodb.com/viz/d612260e-fd12-11e5-9e2b-0ea31932ec1d/public_map Election results 2011 https://benjovy.cartodb.com/viz/479894cc-f84d-11e5-8a89-0ecd1babdde5/public_map Election results 2015 https://benjovy.cartodb.com/viz/dcde4888-f843-11e5-b255-0ecfd53eb7d3/public_map Swing ridings https://benjovy.cartodb.com/viz/3a9cc2f8-fc52-11e5-b294-0ea31932ec1d/public_map Swing_ridings+demographics (manufacturing) https://benjovy.cartodb.com/viz/cd52b6a6-065d-11e6-a5f8-0e674067d321/public_map
I N S Y 4 3 7 M A N A G I N G D AT A & D AT A B A S E
18