The Shift from DB to DC:
Economic Impact on Plan
Presentation to: Workplace Pensions: Next Generation or Final Frontier
Centre for Law in the Contemporary Workplace, Queen’s Law School
Toronto, May 2014
Bob Baldwin
• Terms DB and DC lack precise meanings
• Heterogeneity within each of DB and DC
• DB and DC data may reflect characteristics other than
inherent differences
• EG: size, differences in governance and management, level of
earnings (RPP needs)
• Data not perfectly fit for purpose
• Relevance will trump reliability
Extent of the Shift from DB to DC:
Private Sector 1980
Note too:
• In 2012, 488,000 members in plans transitioning from DB to DC
(compared to 1 m in DC plans)
• PPIC source picks up registered RPPs but not group RRSPs
• In 2011, 3m participate in group RRSPs (3 x the number in registered DC)
• Migration from small registered DC to group RRSP
• Growth in small DB in the private sector
Retirement Wealth
• Participation in an employment pension matters (SFS, 2012)
• Among households aged 55 to 64, 77% have private pension
assets with a median value of $304K
• 69% have RRSPs with a median value of $50K
• Will aim for SFS data on DB vs DC wealth
• MVA per member in DB and DC plans (PPIC)
• In 2012, the average MVA/member in DB plans was $252K vs
$50K for DC plans
• DB plan members more likely to identify their workplace
pension as their primary source of retirement income
• 64% vs 47% (Survey of Financial Preparedness, 2009)
An Accounting Identity
• Contributions + investment returns = benefits + expenses
• Underlines a conceptual point and helps organize information
• Where do investment and other uncertainties show up in DB
and DC
• Pure DB: all on the contribution rate
• Pure DC: all on the benefits
• A reformulation: In DB the rate of saving is adjusted regularly to
meet the benefit obligations; in DC the benefit is adjusted to
reflect the contributions and investment returns
DB vs DC: The General Picture
• Lower level of contributions in DC
• Dollar amounts of contributions per member higher in DB than in registered
DC ($9K vs $3.8K, 2010) – normalized rate data would be preferred
• Actuarial firm identifies 8% as typical DC contribution rate
• Life industry spokesperson: employer GRRSP contributions 4%
• Marginally lower investment returns in DC
• Evidence is from US, sample bias (DC all 100 + members)
• Slight advantage to DB but not consistent from year to year (TW study: 0.76
bps over 17 yrs through 2011; 0.39 in last 5 yrs – net of estimated costs)
• Well documented problems in US 401K investments
• But vs a DC counterfactual, in DB group and individual returns can diverge
• Higher fund management and admin expenses in DC
• Jog to F/P/T ministers on fund management fees: Private DB 30-45 bps; Public
DB 25-35 bps; corp DC 60 bps; corp RRSP 92 bps; retail 40 for ETF/index and
200 on average for active with advice.
• Other sources provide a mixed picture
• Large size (to a point) and capacity for in-house fund management are
clearest determinants of cost
The Predictability Problem with DC
• DC outcomes depend heavily on conditions in financial markets
around retirement age
• Purchases of deferred annuities and life cycle investing can limit the impact –
at a price
• MS modelling
• DC replacement rates vary from a high of 48% (2,000) to 14% (1977):
investment volatility and salary growth
• Blake modelling: 2 scenarios
• Scenario 1: DC portfolio and DB are identical
M result, DC = 100% of DB, 23% chance DC>DB and 44% chance of DC < 25% DB
• Scenario 2 (life cycle): DC portfolio shifts to low risk assets 5 years before
M result, DC = 72 % of DB result, only 7% chance of DC < 25% of DB; only 11%
chance of DC > DB
• NB: mirror image in DB
• Variable contributions (consumption opportunities foregone)
• Amplitude of variation may be smaller in DB (amortization periods,
smoothing of values, etc)
Comfort: An Additional
• No need to worry about
• Investment decisions (low priced access to expertise)
• Longevity risk
• Converting assets from accumulation to decumulation
• But DC is entrenched
• Can be organized to reduce discomfort
Outcomes with No Change to DC
• Impacts greatest for people earning above half AWS
• Greatest need for RPP income to meet earnings replacement
• Impacts on retirement income and retirement age
• Less certainty about amounts of income and retirement age
• Likely lower income and later retirement
• Some attenuation thanks to interactions with other components
of the RIS, income tested benefits and progressive PIT
• Impacts more clear on gross versus net replacement rates
• DB entails variable foregone consumption opportunities prior to
Final Reflections 1
• Future context may be difficult for DB and DC
• Ratio of the retirement period to the period of working life
• Gap between returns on financial assets and salary growth
• But, tighter labour market context may make DB pensions
more interesting for employers
• retention, orderly exit, and mid career recruitment will become
more important
• Pension design is not a binary choice between DB and DC
• Need to explore designs that involve elements of DB and DC –
especially in the private sector
• Predictability of contributions and benefits are both desired
Final Reflections 2
• Single firms are not appropriate organizational platforms for
• Too small, lack expertise, and conflicts
• Even with CPP expansion or ORPP, it is important to get
workplace pensions right
• Next generation or final frontier
• Next generation: pensions as we know them are young; we are
learning how they function in different environments

similar documents