ARGH, bitten by the short edit window again! I wanted to point out that residential meters are measuring kW not kVA. If the pf was .2 and 5kVAh/kWh or 1.0 with 1kVAh/kWh, it wouldn't matter, you'd pay the same per kWh regardless.
Looking more closely, this was a very poor study. They only ran it for 2 months, and didn't have a control. Ironically, it looks like the pf correction caps actually made the pf
worse, because while it would correct a poor pf while the motors were running, it would overcorrect and increase current while the motors were not.
Their conclusion was that the poco may recoup savings on the pf cap due to reduced overall transformer, generator and line current loading, but an individual HO would not see any cost savings.