significantly higher median contributions as a percentage of cash flows from
operations (CONTRATIO) relative to non-freeze firms. These results are consistent
with freeze firms reacting to constrained cash flows and higher funding costs. The
profitability of freeze firms, both on a stand-alone basis (AVG_ROA and ROA) as well
as relative to their industries (ROADIFF ), is also significantly lower, which suggests
that reduced performance levels leading to decreased competitive positions may
motivate the freeze decision. We also find that freeze firms have significantly lower
CEOTENURE compared to non-freeze firms. This is consistent with the possibility
that it is less difficult for a newer CEO to freeze a DB plan than it is for a CEO who is
more deeply entrenched with the firm’s employees. In addition, as expected, freeze
firms are less likely to be unionized. There is no significant difference between freeze
and non-freeze firms with respect to the book-to-market ratio or leverage. Finally, as
our sample is size-matched, it is not surprising that we find no significant difference in
this variable between the two groups[24].
Table IV provides the correlation structure of the variables used in our study. An
interesting observation is that there is a strong positive correlation between
SFAS158_EFFECT and NETPENASSET, which is consistent with the notion that
firms with poorer funded situations face a larger potential impact from SFAS 158. It is
apparent, though not surprising, that many of our independent variables are highly
correlated. Therefore, in conducting multivariate analyses, we first examine
parsimonious models before using more comprehensive specifications. We also use
diagnostic tests including variance inflation factors and condition indices to ensure
that interpretation of results is not problematic.
4.2 Empirical results
Table V presents empirical results on the association of pension accounting balance
sheet variables with firms’ decisions to freeze their DB plans. All models are highly
significant in terms of fit. Consistent with H1 and with our univariate analyses, we find
strong evidence that firms with poorer funded statuses are more likely to freeze their
DB plans, as seen from the highly significant and negative coefficient estimate on
NETPENASSET in model 1a ( p-value ¼ 0.00). We next decompose the funded status
into its constituent elements of PBO and FVA (model 1b), and find similarly significant
results for each component variable[25]. These results are also consistent with H1,
suggesting that firms with higher (lower) pension liabilities (assets) are more likely to
make the decision to freeze their DB plans.
More importantly, to capture the potential impact fromchanges in pension accounting
under FRS 17 and SFAS 158, both of which require recognition of a firm’s funded status
on the balance sheet, we disaggregate the funded status into the amount recognized on
the balance sheet prior to the issuance of SFAS 158 ( BS_RECOG) and the expected
incremental balance sheet impact of SFAS 158 (SFAS158_EFFECT) once it becomes
effective. In model 1c, we find that both BS_RECOG and SFAS158_EFFECT have
significantly positive coefficient estimates ( p-values ¼ 0.00 and 0.00, respectively).
Clearly, the potential impact of SFAS 158 is highly associated with the pension freeze
decision. Results, therefore, remain consistent with H1 suggesting that firms with larger
off-balance sheet liabilities prior to the issuance of SFAS 158 are more likely to make the
pension freeze decision. On the whole, results in Table Vrobustly indicate that firms with
pension plans that are better funded are less likely to freeze their DB plans. Our evidence
is also strongly consistent with firms making the freeze decision to mitigate the probable
balance sheet effects fromexpected pension accounting changes under SFAS 158[26].