Latin American Labor Market Analysis
Research Contributed to the United Nations Chile Data Science Pilot
Explored how digital trace data can be used to inform labor market policies and how job market sites can disclose labor market dynamics, inform about skills and skill-gaps, and present mismatches between supply and demand
Web Scraping Workana
The first part of this analysis was to scrape the website ‘www.workana.org’ and to gather data into a .csv file. Certain job sites are harder to scrape data from so Workana was chosen because not only was it was easier to scrape the necessary data but also because it is a job site primarily for Latin American countries, which is the target of this data science pilot.
The raw .csv file from web scraping the Workana website
Filtering Dataframes
Once the website was scraped and I had a .csv file, I used Jupyter and python to filter the dataframes and I found that many countries had no recorded data or very little recorded data. I decided to filter for only the top five countries with the most amounts of recorded data and used this information for the rest of the analysis.
These countries were: Argentina, Brasil, Colombia, Mexico and Venezuela.
Filtered dataframes to find countries with the most amount of data
Web Design Languages by Country: Pie Chart
The first task I was given in this analysis was to look at the breakdown of the top web design languages within the top five countries using pie charts and heat maps to look at web design skills and skill gaps. The pie charts below clearly displayed the proportion of each web design language within each country. What was found is that most freelancers within Latin American countries have roughly the same proportion of the top five web design languages and a majority of the freelancers knew HTML and JavaScript as their top languages. These languages would be the most relatable and profitable languages to know for people who do freelance work.
Below are the breakdown of the pie charts:
Web Design Languages by Number of Freelancers: Heat Maps
Another part of this first task in the analysis was to look at the breakdown of the top web design languages using heat maps. While the pie charts above show the proportion of each language relative to each country, it is not able to show us the information for all the countries side by side. Heat maps allows us to see which countries have the most number of people who know each web design language and displays this in a matrix.
Below are the breakdowns of the heat maps:
The first heat map shows how Brazil has the most amount of freelance workers using Workana in relation to the other four countries. However, because the distribution of data is uneven amongst the countries, it would be more useful to have a heat map that compares the ratios between the number of freelancers and the populations of the countries.
The second heat map compares the ratios between the number of freelancers and the populations of the countries in order to account for the uneven distribution of data amongst the countries. One can see that HTML and JavaScript are the most popular web design languages amongst all the countries.
Number of Projects Completed vs. Quality of Freelance Work using Scatter Plots
The second task I was given in this analysis was to see if there was a correlation between the number of projects completed versus the quality of the work of the freelancers on Workana using a scatterplot. There were also freelancers who completed more than a thousand projects however these outliers were filtered out. This scatterplot focuses on freelancers who completed between 0-250 projects where the quality ranges from 0-5.
Most of the people who have over 100 projects completed all have ratings between 4.0-5.0 and many people who have less than 50 projects completed have ratings between 0.0-5.0.
The trend seems to be that if a freelancer completes more projects they are more likely to get a higher rating, and this may be due to the fact that experience from the previous projects leads to better freelance work in the future.