Immunization portfolio Convexity/Dispersion

Regarding multiple liabilities dur-matching immunization:
I get that we should:

  1. eliminate the portfolio with lower convexity than that of the liability
  2. select the portfolio with the lowest convexity among the rest of the portfolios, all else equal (min structural risk)

this makes sense to me if we have no view on the expected interest rate change, convexity is good

However my question is, if we are expecting a curve steepening (i.e long end goes up more), a portfolio with higher convexity would be expected to lose more MV than the liability, why would we still want to have a higher convexity for the immunization portfolio?
Maybe my understanding of “immunization” is incorrect…

The immunization you’re discussing is for immunization against a parallel yield curve shift.

If you want to immunize a series of liabilities against structural changes (flattening, steepening, humping, whatever), then you will need to match key-rate durations at a number of maturities along the yield curve. That’s beyond the scope of this reading.

What you described makes a lot of sense to me, however this answer key from the CFAI doesn’t…and it’s been bugging me all day (maybe it’s time to move on)

it’s the “Shrewsbury Case Scenario”

I agree with you; they’re trying to use parallel-shift measures (duration & convexity) to assess a non-parallel shift scenario.

You’re right, they’re wrong; time to move on.