AI Analysis: Railway industry operating statistics of regional companies

Executive Summary

Statistics Canada's Table 23100056 tracks 19 operating statistics for regional railway companies across 38 years (1986–2023), revealing that freight operations overwhelmingly dominate the industry with near-perfect correlations (r ≥ 0.989) across freight-related metrics. The dataset's 652 records show a heavily right-skewed distribution — mean of ~57 million versus a median of ~4 million — driven by total freight car-kilometres reaching a maximum of nearly 660 million. Notably, no statistical outliers were detected across the entire dataset, indicating remarkably stable and consistent reporting over four decades.

Key Findings

• Total freight car-kilometres is the largest operating metric by far, with a maximum of 659.6 million and a median of 554.9 million — dwarfing all other statistics in the dataset.
• The data distribution is heavily right-skewed, with a mean of ~57 million that is more than 14 times the median of ~4 million, driven by a small number of very large freight-related values.
• Freight operations dominate virtually every major metric: total train hours and freight train hours correlate at r = 0.999, and freight train-kilometres account for nearly all total train-kilometres (r = 0.989).
• Loaded freight car-kilometres and total freight car-kilometres share an r = 0.998 correlation, meaning loaded cars account for the vast majority of all car-kilometres travelled by regional railways.
• No outliers were detected across all 652 records using the strict 3x IQR method, indicating highly consistent operational trends and reporting practices throughout the 1986–2023 period.
• Passenger and train-kilometre statistics are substantially smaller in scale than freight metrics, with maximum values in the 6–12 million range compared to hundreds of millions for freight categories.
• The high intercorrelation among freight metrics (loaded car-km, empty car-km, locomotive unit-km, and train hours all at r ≥ 0.992) means these variables are largely redundant, and a single freight indicator could effectively represent the group in any predictive modeling.