Discussion & Conclusion

Interpretation, Limitations, and Concluding Statements.


SES Quintiles and Campaign Instigators

Crowdfunding campaigners that use GoFundMe for cancer-related purposes lie within middle to high levels (approximately 65.49% of campaigns are located in FSAs among the 3 highest quintiles within the income category). With education, the bulk of the campaigners leaned towards the higher quintiles brackets and the middle brackets for housing ownership (see Tables in full report). We cannot assume whether these variables directly influence one’s choice to crowdfund for medical purposes as they could simply a function of population.

Urban and Rural Divide

We also forecasted distinct differences in results between urban and rural areas (J. Snyder & V. Crooks, personal communication, January 25, 2018. This was assumed due to the likelihood of campaigners having of smaller social support networks, lower education levels, and lesser access to technology (J. Snyder & V. Crooks, personal communication, January 25, 2018). Due to the large aggregations of FSAs, it was difficult to determine whether there was a clear urban and rural divide. Upon examination of our visual results (see Figures 7 & 8 in the full report), it is evident that the frequency of campaigns was typically greater closer to city centres. Our results are inconclusive as to whether or not it demonstrates a clear urban or rural divide.

Terms Obtained from Frequent Text Mining

Our results from the text mining aspect of our project can be considered as inconclusive or producing insignificant results. We found that frequent terms in campaign titles and campaign descriptions differed (see full report for figures). The term ‘cancer’ was the most frequently used term at a national scale, which was to be expected as the dataset used was centered on cancer. Verbs such as ‘battle’, ‘fight’, and ‘help’ were frequently seen in campaign titles. Terms including ‘help’, ‘treatment’, ‘family’, and time-related words (i.e. ‘weeks’, ‘months’) are more frequent in crowdfunding campaign descriptions. We did not find any distinct geographic patterns in specific frequency of words used in either campaign titles or descriptions and the terms used were typically similar at both regional and national scales.

Limitations of the Data

There are several limitations that we encountered over the course of this project:

ADA to FSA Conversion Process

Our research is limited by the data collected from GoFundMe. The primary spatial information was the first three digits of the postal codes which were then linked to FSAs. FSAs are smaller in urban areas and may be geographically vast in rural areas. Linking ADAs to FSAs can result in broad generalizations and aggregation errors (N. Schuurman, personal communication, February 19, 2018). This potentially introduced ecological fallacies into our results which this implies that findings from our study are limited, as only broad claims can be stated.

Temporal Attributes of Campaign Data

The data is also only a snapshot that does not capture the dynamic nature of campaigns. It was collected over a period of a few days and only captured the campaigns that existed during that time interval. It also only depicts the campaign state during that specific range, which means that the variables, such as the amount of money collected, are not likely to be accurate representations of the actual value raised. It would be advantageous to collect the campaign data over a longer duration to capture varying states of each campaign as attributes such as the amount raised are not static features. These limitations are unavoidable with the data provided, impacting the selection of methods for this work. The methods were picked to minimize fallacies introduced to the research. Problems arising from the current dataset will help provide direction for future data collection and research.

Quality of Self-Reported Campaigner Information

Another data limitation is that campaigner information is self-authored by persons seeking financial support. Campaigners may craft campaigns with falsified information to motivate donors. Campaigner information is also limited as it contains minimal geographic information besides the city and FSA. Within a campaign description, other details might exist inconsistently throughout the site, such as gender, marital status, and parental status. The success of the campaign is dependent on one’s ability to market themselves and increase the likeliness of donors to take action. Whether campaigner’s truths or embellishments motivate these donors should be assessed in future medical crowdfunding research. Additionally, instances exist in the current dataset where the recipient may be different from campaign’s creator. This poses an issue and necessitates further research because campaigners may reside in or report different FSAs than the individual or family in need of financial assistance. Our analysis was conducted under the assumption that the recipient and campaign creator are the same person or that they live in close proximity to each other (i.e. not in different provinces or FSAs). The postal codes are also self-reported which allows campaign instigators to make errors or misstate their geographic location.

Other Limitations

Several additional limitations were encountered throughout this project. The most prominent is associated with the accuracy of the medical crowdfunding dataset provided at the outset of the project. It was observed that an FSA encapsulating a relatively rural region in southwest of British Columbia (The postal code beginning with V0S) contained an abnormally high number of campaigns (102). Upon closer inspection, we found that this was an error that was likely introduced either when the medical crowdfunding dataset was initially created or by user error. FSAs encoded in the campaign entries dictated the primary level of aggregation for our study, thus necessitating us to limit selection of socioeconomic variables. If spatial and non-spatial attributes were improved, more socioeconomic status variables could be involved, such as gender or immigration status.

Another limitation encountered was the fact that the census profiles at the FSA-level were unavailable during the course of our project due to accuracy issues (Statistics Canada, 2018). To substitute, a methodology to link and weight the chosen socioeconomic variables from the ADA-level to FSA-level had to be developed. Lastly, due to technological constraints, the geographic visual representation of the FSA dataset was simplified for viewing on our web map as its initial file size was too large for the existing configuration. This simplification introduced visual generalizations on our web map but did not impede previously computed numerical results.


The lack of academic literature on medical crowdfunding inspired Snyder et al. (2016) to make a call for further research in the subject matter. Using the data our contact provided, we developed a basic framework that can handle datasets of this nature and can be used to further geographic research on medical crowdfunding. Exploratory work performed enabled regional comparisons using socioeconomic variables related to income, education and housing. We found that the prairie provinces, particularly Alberta and potentially Saskatchewan, could be areas of interest for future studies exploring medical crowdfunding in Canada. These provinces possess unusually high amounts of campaigns relative to their respective populations and to the rest of the nation. Results from the text mining procedure were insignificant and did not expose major geographic trends of discourse utilized across the nation. Definitive claims from results obtained could not be made as it was uncertain whether particular variables were behaving as a function of population.

Due to the above limitations, the robustness of our study was restricted. Future studies of this nature can be improved or yield more significant results if:

  • Future medical crowdfunding dataset contained up-to-date entries obtained at the 6-digit postal code level rather than the 3-digit FSA level.
  • Census profiles for the appropriate census division levels could be obtained rather than using an approach that portioned out statistics weighted based on area.
  • Included some form of exploratory regression work that involved basic socioeconomic variables.