With respect to the control variables in Table V, we use the pension-specific cash flow
variable, CONTRATIO, to investigate how cash flows that directly relate to the current
economic health of the DB plan might motivate the decision to freeze. Alternatively, we
also employ various other pension-specific and overall cash flow measures to control
for the primary funding implications associated with the economic environment and
with the possible apprehension regarding expected changes in funding regulation at
the time[27]. Contrary to our expectation, we find that cash flow motivations are not
significantly associated with the pension freeze decision. Our unexpected findings may
result from the fact that all DB plan sponsors face the challenge of larger and more
volatile contributions depending on economic fluctuations to some degree, which is a
difficulty inherent to DB plan funding. In addition, while such overarching economic
concerns are universal for DB plan sponsors, the ‘‘perfect storm’’ of economic factors
discussed in Munnell and Soto (2007) had a significant impact on aggregate US
pension plan economic status in 2002. Since that time, the DB plan funded status for
most firms has steadily and significantly improved, potentially resulting in a less
volatile and more relaxed pension financing environment, particularly given that legal
funding requirements do not change during our sample period[28].
We expect that a firm’s pension freeze decision may also be based in part on potential
future contributions to a firm’s DB plan, which are partially captured by its PBO and its
related components including assumptions such as settlement rates, increases in salary
levels, etc. Accordingly, it is possible that current cash flows, DB plan contributions,
pension benefits paid, and actual returns to plan assets do not predict potentially significant
increases in expected future plan contributions well.While we do not have a comprehensive
model of determinants of future firm contributions to DB plans, these are likely based not
just on additional pension liability accruals and payouts, but also on factors such as PBGC
requirements, tax regulation, and the actual return on pension plan assets.
We use the firm’s average return on assets for three years prior to the pension freeze
decision (AVG_ROA) to proxy for a firm’s competitive position. Consistent with our
expectation, we find that the estimated coefficient on AVG_ROA is negative and highly
significant ( p-value ¼ 0.00). This indicates that firms with lower profitability are more
likely to freeze their DB plans[29]. We also use an alternative proxy for a firm’s
competitive position, ROADIFF. However, untabulated results show that ROADIFF is
not significant inmotivating the freeze decision. Our evidence is generally consistentwith
claims by management regarding the freeze decision being motivated by deteriorating
profitability impacting the ability to compete effectively. The results also imply that poor
financial performance generallymaymotivate a firmto freeze its DB plan.
Of the remaining control variables, CEOTENURE is significant and negative,
consistent with our expectation that a newer CEO is more likely to have both the
incentive and the political clout to freeze a DB plan as a perceived way of improving a
firm’s financial performance and reducing risk. The UNION dummy variable is also
negative and significant indicating that the presence of a labor union reduces the
probability of a pension freeze. The fact that LEVand BMRATIO are insignificant could
be an artifact of our matching process as matching by size and industry could have
controlled for these factors as well. Their statistical insignificance in our models
therefore does not necessarily signify that these factors are unimportant in firms’
consideration of the pension freeze decisions.
Overall, our evidence is consistent with the notion that financial reporting incentives
and expected changes in accounting standards, particularly those incorporated in SFAS
158, significantly and incrementally contribute to the DB plan freeze decision[30].
Though we cannot conclude that the impending issuance of SFAS 158 alone caused
firms to freeze their DB plans, we can say that we present a powerful link between
potential changes in pension accounting standards and the DB plan freeze decision,
while controlling for other motivations.We provide evidence confirming a major increase
in incidence of DB plan freezes in the years subsequent to the notable issuance of FRS 17
in the UK and preceding the issuance of SFAS 158 in the USA. Additionally, we provide
robust evidence of a strong association between the direct balance sheet reporting
impact of SFAS 158 and the freeze decision. Thus, after controlling for alternative
incentives including cash flow, competitive pressure, and industry effects, our results
indicate that firms likely to be most significantly affected by a full fair-value pension
accounting standard (e.g. FRS 17 or SFAS 158) appear to have timed their DB plan freeze
decisions in anticipation of the issuance of such a standard.
4.3 Supplemental analyses
4.3.1 Analysis of pension accounting income statement variables. In Table VI we
present empirical results from supplemental tests on the association of firms’ decisions
to freeze their DB plans with pension accounting income statement variables. Each
model is significant in terms of fit. However, we do not find a significant association
between the net periodic pension cost variable, AVERAGEPENCOST, and the decision
to freeze. Therefore, model 2a is not consistent with an expectation that a higher net
periodic pension cost motivates firms to freeze their DB plans. In addition, it is
apparent based on variance inflation factor and condition index diagnostics that
multicollinearity influences the precision of the coefficient estimates of our individual
pension cost component variables, which detracts from our ability to identify
associations in model 2b. Though our results do indicate that one of the pension cost
component variables, AVERAGEASSETRET, is significantly associated with the
freeze decision ( p-value ¼ 0.04), the lack of precision in its coefficient estimate due to
multicollinearity issues impairs our ability to interpret the result in a meaningful
way[31]. Therefore, we cannot reject the null hypothesis of a significant association
between our pension income statement variables and the DB plan freeze decision.
Results related to the control variables, however, are consistent with our findings in
Table V. Our cash flow variable, CONTRATIO, is not significantly associated with the
pension freeze decision. Consistent with our expectation, we find that the estimated
coefficient on AVG_ADJ_ROA, which is the three year historical average ROA
adjusted for pension expense, is negative and highly significant ( p-value ¼ 0.00). This
indicates that firms with lower profitability relative to assets are more likely to freeze
their DB plans. Also consistently significant across model 2 regressions are
CEOTENURE and UNION, which supports results in Table V suggesting that firms
with entrenched CEOs and labor unions are less likely to freeze their DB plans.
4.3.2 Analysis of freeze type (hard vs soft/partial). While we do not distinguish
between the different types of DB plan freezes in our main tests, we do performadditional
analyses in an effort to determine the impact of firms’ financial accounting, cash flow, and
competitive pressure characteristics on decisions concerning DB plan freeze type (i.e.
hard or soft/partial). Untabulated univariate results reveal that none of our independent
variable means and medians are significantly different based on freeze type. This
suggests that concerns related to financial accounting, cash flow, and competitive
pressures do not significantly influence the type of freeze that a firm chooses.
To get a clearer picture of how certain firm characteristics might motivate firms to
choose one particular type of freeze over another, we also perform a multinomial