AI Analysis: Reason for leaving job during previous year, monthly, unadjusted for seasonality

Executive Summary

Statistics Canada's Table 14100125 tracks monthly job separations by reason across nearly 50 years (1976–2026), covering 161,263 records across 13 departure categories, 11 regions, 7 age groups, and 3 gender categories. The data is strongly right-skewed, with a median of 7.3K persons versus a mean of 71.2K, driven by large aggregate totals reaching nearly 8 million persons. Early trend data (1976–1978) reveals a notable shift toward involuntary job loss, with permanent layoffs rising from 707K to 916K while voluntary job leavers declined from 1,456K to 1,264K.

Key Findings

• The dataset contains 161,263 non-null records spanning January 1976 to February 2026, with 9,009 unique time series vectors reflecting a highly granular combination of regions, age groups, genders, and departure reasons.
• The distribution is strongly right-skewed: the median value is just 7.3K persons while the mean is 71.2K — nearly 10x higher — with a maximum of 7,998.5K persons representing broad aggregate categories.
• Involuntary job loss grew significantly in the early data period, with job losers rising from 804K to 993K and permanent layoffs increasing from 707K to 916K between January 1976 and April 1978.
• Voluntary departures (job leavers) declined from 1,456K to 1,264K over the same 1976–1978 period, suggesting a weakening labour market where workers had less confidence to quit voluntarily.
• 'Have not worked in last year' was the second-largest category and grew from 3,446K to 3,794K between 1976 and 1978, pointing to a rising share of people detaching from the workforce.
• No statistical outliers were detected across all 13 job-leaving categories using a 2.5× IQR threshold, suggesting that even major economic shocks like recessions or COVID-19 produced gradual rather than abrupt shifts within the long historical range.
• The 50% interquartile range of values falls between just 2.1K and 31.2K persons, indicating that the vast majority of granular subgroup observations are relatively small, with extreme values confined to high-level aggregate totals.