Separate historical counts instead of inventing a rate

Question. How did unemployment and the armed forces move against employment?

Encoding. Two independently labeled historical count paths keep distinct populations separate.
Ordering. Calendar years in both panels.
Baseline. Both value scales include zero; these are counts, not a computed labor-force share.
Scale. Shared time domain with separate count domains, explicitly not a stacked total or dual-axis overlay.

What this design gives up

Comparing panel heights without reading the scales is invalid. No denominator for an official unemployment rate is supplied.

The selected columns do not provide a verified labor-force denominator for constructing a standard unemployment rate. Population is not a substitute. These source counts belong to an old benchmark vintage.