Social conditions: Stats lab
How do poverty, housing and food insecurity move together?
Put any two provincial indicators on the axes, fit a line, and judge what the relationship can and cannot tell you. With ten provinces, this is also a lesson in how fragile small samples are.
The lab
Food insecurity against Poverty rate (MBM)
Each dot is a province, sized by population. Line: ordinary least squares fit, unweighted.
Food insecurity = 13.91 + 1.047 × Poverty rate (MBM)Slope standard error: 0.53. With so few observations, one province can change the result: toggle the territories and watch.
Method notes
Exercises
With the defaults (poverty on x, food insecurity on y), report r, R² and the slope. Interpret the slope in plain words.
Show a hint
The slope tells you how many percentage points of food insecurity go with one more point of poverty, across provinces.
Turn on the territories. What happens to the slope and R²? Which territory drives the change, and why might it be different?
Check your reasoning
Nunavut, with very high poverty and food insecurity, pulls the line steeply upward and raises R². Its costs of food and housing, remoteness and demographic profile differ greatly from the provinces, so it may follow a different relationship.
Plot food insecurity against median after-tax income. Then against core housing need. Which explains more of the variation, and what does that suggest about the role of housing costs?
Check your reasoning
Compare the R² values. If housing need explains more than income, it suggests that what is left after rent matters more than income itself, consistent with research that renters and high-cost cities face more food insecurity.
Download the CSV and fit a multiple regression of food insecurity on poverty and core housing need in Python. Do both stay significant? Discuss collinearity.
Show a hint
import statsmodels.formula.api as smf; smf.ols('food_insecurity_pct_2024 ~ poverty_pct_2024 + core_housing_need_pct_2022', data=df).fit().summary()