Page 183 - 35-1

NTU Management Review Vol. 35 No. 1 Apr. 2025

Both datasets are publicly available on their respective websites. Financial metrics,
including SPIs and other data, are obtained from Compustat and CRSP. The final sample
comprises 296,733 firm-quarter observations for testing H1. Owing to the requirement of
voluntary non-GAAP earnings disclosures, the observations for testing H2 are 44,632.

To test H1, we employ the following logistic regression model (1):

Pr (Ngp_Dis = 1)
it
= G (β + β EPU + β Spi + β EPU × Spi + Σβ Controls + Fixed Effects). (1)
it
0
3
t
1
2
t
it
n
it
The subscripts i and t of each variable in all models denote the firm and the quarter of
the year for an observation, respectively. G(.) is a logistic function. Ngp_Dis is an indicator
variable that equals one if the firm voluntarily discloses non-GAAP earnings information
in the quarter and zero otherwise. Spi is an indicator variable that equals one if the firm
recognizes SPIs in the quarter and zero otherwise. The EPU index (EPU) is originally
developed by Baker et al. (2016). In accordance with Nagar et al. (2019), we further adjust
the index by taking the natural logarithm of the average of the monthly EPU index over
the quarter.
To test H2, we utilize the following ordinary least squares (OLS) regression model (2):

Future_OI it

= λ + λ EPU + λ Ngp_Exclu + λ EPU × Ngp_Exclu + Σλ Controls
t
3
it
it
n
it
0
t
1
2
+ Fixed Effects + ε . (2)
it
Building on prior empirical studies (e.g., Curtis et al., 2014; Chen, Lee, Lo, and Yu,
2021), we utilize the predictive ability of the excluded items for future operating income
(Future_OI) to measure the exclusion quality of non-GAAP earnings. Future_OI is four-
quarter-ahead operating income, scaled by total assets at the end of the quarter. Ngp_
Exclu denotes the total excluded items, calculated as the firm’s non-GAAP earnings minus
GAAP earnings, scaled by total assets at the beginning of the quarter. A positive Ngp_
Exclu indicates that the net value of total excluded items is negative in terms of GAAP
earnings. Thus, a positive coefficient on Ngp_Exclu in model (2) implies a high exclusion

quality.

175