• No se han encontrado resultados

Información riesgos y medidas preventivas

The descriptive statistics are presented in Table 4.3. All continuous variables are winsorized at the top and bottom 1%. Following Banker et al. (2017), net income (𝐸𝐴𝑅𝑁𝑖𝑡), operating cash flow change (𝛥𝐶𝐹𝑖𝑡) and sales change (𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡) are scaled by the market value of equity at the beginning of the year. On average, net income is equal to −0.7% of lagged market value, and the median is 3%, which is at a similar level to prior studies (e.g., Banker et al. 2017). Consistent with the presence of conditional conservatism (Ball et al., 2000; Banker et al., 2017; Basu, 1997), scaled earnings are negatively skewed with a mean smaller than the median. Annual stock returns for O&G sector is 9.87% on average and the median is 5.32%. Some 45.18% of the sample has a negative stock return with 𝐷𝑅𝑖𝑡 equal to 1, which is similar to the market- wide stock return performance documented in Banker et al. (2017). Average operating cash

decreases (𝐷𝐶𝑖𝑡) and sales decreases (𝐷𝑆𝑖𝑡) account for 46.49% and 40.13% of the observations, respectively. In general, the distribution of net income, stock returns, operating cash flow change and sales change for this sample is largely consistent with prior studies (e.g., Banker et al., 2017). Annual crude oil price return is 7.12% for the period 2001–2015 and the median is 4.25%.63 The positively skewed oil price return indicates an overall increasing trend for oil prices during the sample period.

Table 4.4 reports the Pearson correlations among variables. The pairwise correlation coefficients between stock return (𝑅𝐸𝑇𝑖𝑡), scaled cash flow change (𝛥𝐶𝐹𝑖𝑡) and scaled sales change (𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡) are 0.1785 and 0.1071 respectively. The relatively low correlation coefficients indicate that these three indicators capture different aspects of firm performance and provide complementary information. The correlation coefficient between 𝛥𝐶𝐹𝑖𝑡 and

𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 is 0.4974 (significant at the 0.05 level, two-tailed), suggesting a stronger association between cash flow and sales for the O&G sector compared with market-wide statistics (0.169) shown in Banker et al. (2017). These three indicators are significantly positively correlated with earnings. 𝑂𝑃𝑅𝑖𝑡−1 is positively correlated with earnings, operating cash flow changes and sales changes with coefficients of 0.2272, 0.2987 and 0.4021 respectively, consistent with the argument that oil price changes have a positive influence for O&G firms performance.64

63 Lagged oil price returns for the sample period 2002–2016 are actually the returns for 2001–2015. The mean and

median for the oil price returns for 2002–2016 are 0.3481 and 0.1094 respectively. The differences in average returns for these two period is mainly attributed to the oil price recovery in 2016.

64 Though the correlation coefficient between 𝑂𝑃𝑅

𝑖𝑡−1 and 𝑅𝐸𝑇𝑖𝑡 is not statistically significant, 𝑅𝐸𝑇𝑖𝑡 is

positively correlated with concurrent oil price returns with a coefficient of 0.3843 (significant at the 0.05 level, two-tailed), indicating that stock prices react faster (than a one-year lag) to oil price changes.

Table 4.3 Descriptive Statistics for Full Sample

Variable Mean S.D. Min Q1 Median Q3 Max

𝐸𝐴𝑅𝑁𝑖𝑡 -0.0726 0.3531 -2.0432 -0.0801 0.0305 0.0834 0.3972 𝐷𝑅𝑖𝑡 0.4518 0.4979 0.0000 0.0000 0.0000 1.0000 1.0000 𝑅𝐸𝑇𝑖𝑡 0.0987 0.5432 -0.8696 -0.2656 0.0532 0.4001 2.0955 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 -0.1586 0.2375 -0.8696 -0.2656 0.0000 0.0000 0.0000 𝐷𝐶𝑖𝑡 0.4649 0.499 0.0000 0.0000 0.0000 1.0000 1.0000 𝛥𝐶𝐹𝑖𝑡 -0.0206 0.1672 -0.7341 -0.0618 0.0074 0.0529 0.4635 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 -0.0614 0.1316 -0.7341 -0.0618 0.0000 0.0000 0.0000 𝐷𝑆𝑖𝑡 0.3913 0.4882 0.0000 0.0000 0.0000 1.0000 1.0000 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 -0.0216 0.2833 -1.3574 -0.0646 0.0281 0.0984 0.6694 𝐷𝑆𝑖𝑡×𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 -0.0941 0.2302 -1.3574 -0.0646 0.0000 0.0000 0.0000 𝐷𝑂𝑖𝑡−1 0.4281 0.4950 0.0000 0.0000 0.0000 1.0000 1.0000 𝑂𝑃𝑅𝑖𝑡−1 0.0712 0.3366 -0.5052 -0.2114 0.0425 0.2301 1.0096 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 -0.0952 0.1476 -0.5052 -0.2114 0.0000 0.0000 0.0000 𝑊𝐷𝑖𝑡 -0.0035 0.0154 -0.1096 0.0000 0.0000 0.0000 0.0000 𝐺𝑊𝑖𝑡 -0.0067 0.036 -0.2765 0.0000 0.0000 0.0000 0.0000

The table reports descriptive statistics for 1,224 firm-year observations from 2002–2016. All continuous variables are winsorized at the top and bottom 1%.

Table 4.4 Correlation Matrix EARN DR RET DR× RET DC ΔCF DC× ΔCF DS ΔSALES DS× ΔSALES DO (t-1) OPR (t-1) DO× OPR (t-1) WD GW 𝐸𝐴𝑅𝑁𝑖𝑡 1 𝐷𝑅𝑖𝑡 -0.2216* 1 𝑅𝐸𝑇𝑖𝑡 0.2040* -0.7386* 1 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 0.3961* -0.7372* 0.7385* 1 𝐷𝐶𝑖𝑡 -0.2959* 0.0952* -0.0935* -0.1141* 1 𝛥𝐶𝐹𝑖𝑡 0.4213* -0.1115* 0.1758* 0.1925* -0.6246* 1 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 0.5263* -0.0864* 0.0676* 0.2078* -0.5209* 0.8649* 1 𝐷𝑆𝑖𝑡 -0.3611* 0.1197* -0.0986* -0.1286* 0.4643* -0.3899* -0.3936* 1 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 0.4254* -0.0978* 0.1071* 0.1590* -0.3534* 0.4974* 0.5334* -0.5849* 1 𝐷𝑆𝑖𝑡×𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 0.4600* -0.0458 0.022 0.1473* -0.2979* 0.4418* 0.5726* -0.4832* 0.9020* 1 𝐷𝑂𝑖𝑡−1 -0.2225* -0.0887* 0.0665* 0.0567* 0.2695* -0.3063* -0.3138* 0.4362* -0.3756* -0.3261* 1 𝑂𝑃𝑅𝑖𝑡−1 0.2272* 0.0013 -0.0091 -0.0144 -0.2709* 0.2987* 0.3036* -0.4442* 0.4021* 0.3475* -0.7549* 1 𝐷𝑂𝑖𝑡−1×𝑂𝑃𝑅𝑖𝑡−1 0.3309* 0.0091 -0.036 0.0488 -0.3768* 0.3617* 0.4013* -0.5602* 0.5006* 0.4738* -0.7457* 0.7576* 1 𝑊𝐷𝑖𝑡 0.2109* -0.0296 0.0257 0.0886* -0.1287* 0.1167* 0.1348* -0.1173* 0.1810* 0.1997* -0.0686* 0.0586* 0.1140* 1 𝐺𝑊𝑖𝑡 0.2789* -0.1353* 0.1519* 0.2470* -0.0408 0.0474 0.0759* -0.0582* 0.1353* 0.1909* -0.0233 -0.0038 0.0909* 0.1671* 1

Pearson correlations are reported for the sample of 1,224 O&G firm-year observations from 2002 to 2016. Correlations with * are statistically significant at the 5 percent level. The variable definitions are provided in Table 4.1.

Empirical Results

Table 4.5 reports regression estimates for Model 4.4. Columns 1 and 2 of Table 4.5 report the Banker et al. (2017) three-indicator model (Model 4.3), with (Column 1) and without (Column 2) controlling for firm fixed effects respectively, for comparison.65 Column 3 reports the

regression results for the extended Banker et al. (2017) model (4.4) with the additional indicator of changes in oil prices. The regression is estimated with control for firm fixed effects, as suggested by Ball et al. (2013). As oil price return is calculated on an annual basis, which already accounts for change in economic conditions across years, year fixed effects are not included in the regressions with variables of oil price returns. Similar to Basu (1997), the asymmetric timeliness coefficient on 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡is positive and significant (at the 0.01 level, two-tailed). The positive coefficient indicates that bad news (negative𝑅𝐸𝑇𝑖𝑡) is recognized in concurrent earnings more fully than good news (positive𝑅𝐸𝑇𝑖𝑡), suggesting conditional conservatism for stock returns. Consistent with Ball and Shivakumar (2006), the coefficient for

𝐷𝐶𝑖𝑡 × 𝛥𝐶𝐹𝑖𝑡 is positive and significant (at the 0.01 level, two-tailed), indicating asymmetric timeliness with respect to concurrent operating cash flow change. The coefficient on 𝐷𝑆𝑖𝑡× 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 is consistent with Banker et al. (2017)—positive and significant—indicating asymmetric timeliness with respect to sales changes.

Table 4.5 Asymmetric Timeliness Estimates for Multiple Indicators Based on Banker et al. (2017) Banker et al. (2017) Full Three- Indicator Model Banker et al. (2017) Full Three- Indicator Model Four-Indicator Model VARIABLES predicted sign (1) EARN (2) EARN (3) EARN 𝐷𝑅𝑖𝑡 0.0392* 0.0330* 0.0381* (1.8618) (1.6498) (1.8007) 𝑅𝐸𝑇𝑖𝑡 -0.0113 -0.0206 -0.0096 (-0.3680) (-0.7404) (-0.3104) 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 + 0.3488*** 0.2911*** 0.3438*** (5.4533) (4.4574) (5.3546) 𝐷𝐶𝑖𝑡 -0.0022 0.0004 0.0009 (-0.1285) (0.0237) (0.0523) 𝛥𝐶𝐹𝑖𝑡 -0.0689 0.0320 -0.0587 (-0.4451) (0.2167) (-0.3769) 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 + 0.7158*** 0.4871** 0.6966*** (3.1213) (2.2604) (3.0179) 𝐷𝑆𝑖𝑡 -0.0747*** -0.0385* -0.0659*** (-4.0139) (-1.9390) (-3.1882) 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 -0.1070 -0.0744 -0.1076 (-1.5487) (-1.0934) (-1.5251) 𝐷𝑆𝑖𝑡×𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 + 0.3424*** 0.2656*** 0.3342*** (3.8131) (3.0811) (3.6025) 𝐷𝑂𝑖𝑡−1 -0.0004 (-0.0260) 𝑂𝑃𝑅𝑖𝑡−1 -0.0094 (-0.3437) 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 + 0.0872 (0.9824)

Firm-fix effect - Yes Yes

Year-fix effect Yes Yes -

Observations 1,224 1,224 1,224

Adjusted R2 0.480 0.545 0.480

F-statistic for the full effect of

𝐷𝑅𝑖𝑡, 𝑅𝐸𝑇𝑖𝑡, 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 18.31*** 8.94*** 18.15***

𝐷𝐶𝑖𝑡, 𝛥𝐶𝐹𝑖𝑡, 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 14.64*** 12.13*** 13.89***

𝐷𝑆𝑖𝑡, 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡, 𝐷𝑆𝑖𝑡×𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 26.43*** 10.57*** 15.97***

𝐷𝑂𝑖𝑡−1, 𝑂𝑃𝑅𝑖𝑡−1, 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 0.57

F-statistics for the asymmetric effect of

𝐷𝑅𝑖𝑡, 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 24.90*** 16.90*** 24.13***

𝐷𝐶𝑖𝑡, 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 9.75*** 5.03** 9.00***

𝐷𝑆𝑖𝑡, 𝐷𝑆𝑖𝑡×𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 19.44*** 10.75*** 16.29***

*** p < 0.01, ** p < 0.05, * p < 0.1

The t-statistics in parentheses are based on standard errors clustered by firm and year. This table presents the results for model:

Column 1 and Column 2:

𝐸𝐴𝑅𝑁𝑖𝑡=𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡+ 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛾1𝐷𝑆𝑖𝑡+

𝛾2𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡+ 𝛾3𝐷𝑆𝑖𝑡× 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡+ 𝜍𝑖𝑡

Column 3:

𝐸𝐴𝑅𝑁𝑖𝑡=𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡+ 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑖𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛾1𝐷𝑆𝑖𝑡+ 𝛾2𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡+ 𝛾3𝐷𝑆𝑖𝑡× 𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡+ 𝛿1𝐷𝑂𝑖𝑡−1+ 𝛿2OPR𝑖𝑡−1+ 𝛿3𝐷𝑂𝑖𝑡× 𝑂𝑃𝑅𝑖𝑡−1+ 𝜍𝑖𝑡

The variable definitions are provided in Table 4.1.

H4.1a predicts that after controlling for the asymmetric effect of stock returns, operating cash flow changes and sales changes, earnings exhibit asymmetric association with oil price changes. The coefficient for 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 is therefore expected to be positive. Column 3 in Table 4.4 reports the regression for the full four-indicator model (Model 4.4), which examines the incremental effect of oil price changes on asymmetric timeliness of earnings recognition after controlling for stock returns, operating cash flow changes and sales changes. The results show that the coefficient for 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 is positive (0.0872) but not significant.

Sales are determined by price and quantity. Given the relatively high correlation between

𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 and𝑂𝑃𝑅𝑖𝑡−1 (0.4021, significant at the 0.05 level) for my sample and high VIF for

𝛥𝑆𝐴𝐿𝐸𝑆𝑖𝑡 (7.61) in regression Model 4.4, the coefficient for 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1could be biased by collinearity issues between sales and price.66 I therefore replace 𝛥𝑆𝐴𝐿𝐸𝑆

𝑖𝑡 in the Banker et al. (2017) model with lagged oil price change (𝑂𝑃𝑅𝑖𝑡−1) to test whether earnings exhibit asymmetric association with oil price changes incremental to stock returns and operating cash flow changes to test for H4.1b. The model is specified as:

𝐸𝐴𝑅𝑁𝑖𝑡 = 𝜕0 + 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡 + 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡+ 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡 + 𝛽3𝐷𝐶𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛿1𝐷𝑂𝑖𝑡−1+ 𝛿2OPR𝑖𝑡−1 + 𝛿3𝐷𝑂𝑖𝑡× 𝑂𝑃𝑅𝑖𝑡−1+ 𝜀𝑖𝑡

where all variables are as defined previously.

Column 1 of Table 4.6 reports the Ball and Shivakumar (2006) model before sales have been added. The results suggest that earnings respond asymmetrically to stock returns and changes in operating cash flows, with statistically significant coefficients for 𝐷𝑅𝑖𝑡 × 𝑅𝐸𝑇𝑖𝑡 (0.3644) and

𝐷𝐶𝑖𝑡 × 𝛥𝐶𝐹𝑖𝑡 (1.2140) (significant at the 0.01 level, two-tailed). When lagged oil price return is added to the model, the coefficient on 𝐷𝐶𝑖𝑡 × 𝛥𝐶𝐹𝑖𝑡 is reduced by 16.22%, from 1.2140 in the Ball and Shivakumar (2006) model (Column 1) to 1.0518 in the three-indicator model with lagged oil price return (Column 2). This suggest that oil price return is an important correlated omitted variable in the Ball and Shivakumar (2006) model for firms in the O&G industry. Column 2 in Table 4.6 reports the regression results for the three-indicator model, with sales changes replaced by lagged oil price returns.67 The coefficient on 𝐷𝑂

𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 is 0.2992,

significant at the 0.01 level (two-tailed), indicating that earnings are more sensitive to bad news (negative 𝐷𝑂𝑖𝑡−1) than good news (positive 𝐷𝑂𝑖𝑡−1), suggesting conditional conservatism for oil price returns.

67 The mean VIF for Model 4.5 is 3.29 and VIFs for variables in the model are all below 5, indicating low likelihood

Table 4.6 Asymmetric Timeliness Estimates for Multiple Indicators Based on Ball and Shivakumar (2006)

Ball and Shivakumar (2006)

Three-Indicator Model: with lagged Oil Price

Return VARIABLES predicted sign (1) EARN (2) EARN 𝐷𝑅𝑖𝑡 0.0377* 0.0372* (1.7293) (1.7173) 𝑅𝐸𝑇𝑖𝑡 -0.0180 -0.0123 (-0.5642) (-0.3916) 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 + 0.3644*** 0.3495*** (5.4376) (5.3065) 𝐷𝐶𝑖𝑡 -0.0353** -0.0107 (-1.9726) (-0.5811) 𝛥𝐶𝐹𝑖𝑡 -0.2465 -0.1835 (-1.5409) (-1.1414) 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 + 1.2140*** 1.0518*** (5.4194) (4.5684) 𝐷𝑂𝑖𝑡−1 0.0104 (0.6562) 𝑂𝑃𝑅𝑖𝑡−1 -0.0042 (-0.1541) 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 + 0.2992*** (3.8476)

Firm-fix effect Yes Yes

Observations 1,224 1,224

Adjusted R2 0.432 0.447

F-statistic for the full effect of

𝐷𝑅𝑖𝑡, 𝑅𝐸𝑇𝑖𝑡, 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 18.11*** 18.15***

𝐷𝐶𝑖𝑡, 𝛥𝐶𝐹𝑖𝑡, 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 45.49*** 28.86***

𝐷𝑂𝑖𝑡−1, 𝑂𝑃𝑅𝑖𝑡−1, 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 10.64***

F-statistics for the asymmetric effect of

𝐷𝑅𝑖𝑡, 𝐷𝑅𝑖𝑡×𝑅𝐸𝑇𝑖𝑡 25.40*** 24.11***

𝐷𝐶𝑖𝑡, 𝐷𝐶𝑖𝑡×𝛥𝐶𝐹𝑖𝑡 30.94*** 20.95***

𝐷𝑂𝑖𝑡−1, 𝐷𝑂𝑖𝑡−1× 𝑂𝑃𝑅𝑖𝑡−1 15.00***

*** p < 0.01, ** p < 0.05, * p < 0.1

The t-statistics in parentheses are based on standard errors clustered by firm and year. This table presents the results for model:

Column 1 : 𝐸𝐴𝑅𝑁𝑖𝑡=𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡+ 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑖𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝜈𝑖𝑡

Column 3: 𝐸𝐴𝑅𝑁𝑖𝑡=𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡+ 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑖𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛿1𝐷𝑂𝑖𝑡−1+ 𝛿2OPR𝑖𝑡−1+ 𝛿3𝐷𝑂𝐼𝑡× 𝑂𝑃𝑅𝑖𝑡−1+ 𝜈𝑖𝑡

a $1 oil price return decrease (bad news) reduces earnings by 29.5 cents on average (=−0.0042 + 0.2992), indicating much quicker recognition of bad news than good news for oil price changes. The F-statistic, which tests whether positive and negative oil price returns have equal influence (𝐷𝑂𝑖𝑡−1= 𝐷𝑂𝑖𝑡−1×𝑂𝑃𝑅𝑖𝑡−1) is 15.00 (significant at the 0.01 level), indicating strong asymmetric response of earnings to oil price returns. Thus, this result supports the prediction that earnings exhibit asymmetric association with oil price changes after controlling for the asymmetric effect of stock return and operating cash flow changes.

Overall, the results indicate that for firms in the O&G industry, earnings respond asymmetrically to good news and bad news from changes in oil prices. Bad news (negative oil price returns) are more fully recognized in concurrent earnings than good news (positive oil price returns). Oil price changes provide complementary information to stock returns and operating cash flow changes, which reflect information on future cash flows, supporting H4.1b. However, in the full four-indicator model (Model 4.4), neither positive nor negative lagged oil price returns exhibit significant association with earnings, because of the correlation between sales changes, oil price returns and earnings. The results suggest that lagged oil price changes and sales changes provide similar information regarding future cash flows.68

Additional Analysis

Banker et al. (2017) suggest that asset write downs are the most fundamental manifestation of conservatism. Impairment tests are conducted for asset groups “at the lowest level for which identifiable cash flows are largely independent of cash flows of other assets and liabilities”

68 The results are consistent with firms as price takers with inelastic demand for their product across a reasonable

price range. Under such assumptions, quantities produced and sold are expected to remain constant and the information provided by sales changes would primarily reflect the change in product market prices.

(SFAS, 144, para. 10). Therefore, indicators that help predict future cash flows for individual asset classes are relevant for assessing impairment. Banker et al. (2017) show that asset write off and impairment losses respond asymmetrically to stock returns, operating cash flow changes and sales changes. This study extends prior literature to test whether product market prices exhibit a similar asymmetric association with asset impairment.

This study argues that oil price changes are expected to be an important factor considered by accountants in their impairment decisions in O&G firms because of the impact on inventory write-downs, long-lived asset write downs, impairment of intangible assets, and measuring and recognizing goodwill from acquisitions. For example, inventories are stated at the lower of cost and net realizable value.69 If oil prices decrease to a level lower than production cost, value of

inventories needs to be impaired to net realizable value (PwC, 2017).70

For long-lived assets, SFAS 144 requires a two-step impairment test. The first step is to compare the estimated sum of undiscounted future cash flow associated with the asset (or asset group) with the asset’s carrying amount. If the undiscounted cash flows are less than the asset’s carrying amount, then in the second step, the asset is written down to its fair value. The fair value is calculated based on market prices if there is an active market for the assets or estimated as the total of discounted future cash flows when quoted market prices are absent. Indicators for decreases in expected future cash flows increase both the probability (in step 1) and the magnitude (in step 2) of the asset impairment. For O&G companies, impairment assessments for long-lived assets incorporate commodity market uncertainties including projected commodity pricing, supply and demand for goods and services, and future market conditions (Halliburton, 2017). If crude oil prices decline significantly and/or remain at low levels for a

sustained period of time, the long-lived assets’ carrying amounts are likely to be impaired because of the decrease in projected future cash flows. Given the long-term nature of many large-scale development projects, even perceptions of longer-term lower oil prices by oil firms can cause them to reduce or defer major expenditure (Halliburton, 2017).71 Decreases in oil prices can affect the overall returns for these projects by either extending the time until the expected returns are realized or recognising impairment expenses for the assets.

Goodwill is recognized when the consideration paid is greater than the fair value of net asset acquired in M&A deals. Goodwill can represent access to new markets, community/government relationship, portfolio management, technology and expertise (PwC, 2017). O&G companies might be inclined to pay a premium to protect the value of existing O&G operations that they already own. As per ASU 2011-08, firms have the option of first performing a qualitative assessment to test goodwill for impairment on a reporting-unit-by- reporting-unit basis.72 If the qualitative assessment indicates that it is probable that carrying amount is greater than the fair value of a reporting unit, the entity will conduct the two-step goodwill impairment test as described in ASC 350; otherwise, the two-step goodwill impairment test is not required. When oil prices increase, firms pay premiums in acquisition deals for the high projected future cash flows that lead to goodwill recognition. When oil prices decrease significantly, estimated fair value based on projected future cash flows can be potentially below carrying amounts, leading to impairment of goodwill.

71 Halliburton is the eighth largest U.S. oil company based on market value (40.49 billion USD) in 2018

(www.statistc.com). Impairment for Halliburton in the 2016 fiscal year was 3.36 billion USD, whereas impairment

for 2017 was 647 million USD, indicating a reverse relation between oil prices and corporate asset impairment recognitions. Halliburton suggests that the impairment in 2016 was primarily a result of the down turn in the energy market. The impairment consisted of fixed asset impairment and write offs, inventory write-downs, and impairment of intangible assets.

72 For example, in Halliburton, the test for goodwill impairment is carried out for each reporting unit, which is the

Overall, accounting standards for asset write downs and impairment suggest that firms will recognize unrealized losses reflected in unfavorable indicators of future cash flows in a timely manner. To test whether asset write downs and goodwill impairment exhibit an asymmetric association with oil price changes, following Banker et al. (2017), I replace the dependent variable, 𝐸𝐴𝑅𝑁𝑖𝑡, with long-lived asset write down (𝑊𝐷𝑖𝑡) and goodwill impairment (𝐺𝑊𝑖𝑡) in Model 4.5,73 to formulate the following two models:

𝑊𝐷𝑖𝑡 = 𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡 + 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛿1𝐷𝑂𝑖𝑡−1 + 𝛿2OPR𝑖𝑡−1+ 𝛿3𝐷𝑂𝑖𝑡× 𝑂𝑃𝑅𝑖𝑡−1+ 𝜀𝑖𝑡 (4.6) 𝐺𝑊𝑖𝑡 = 𝜕0+ 𝜕1𝐷𝑅𝑖𝑡+ 𝜕2𝑅𝐸𝑇𝑖𝑡+ 𝜕3𝐷𝑅𝑖𝑡× 𝑅𝐸𝑇𝑖𝑡 + 𝛽1𝐷𝐶𝑖𝑡+ 𝛽2𝛥𝐶𝐹𝑖𝑡+ 𝛽3𝐷𝐶𝑡× 𝛥𝐶𝐹𝑖𝑡+ 𝛿1𝐷𝑂𝑖𝑡−1 + 𝛿2OPR𝑖𝑡−1+ 𝛿3𝐷𝑂𝑖𝑡× 𝑂𝑃𝑅𝑖𝑡−1+ 𝜀𝑖𝑡 (4.7)

where 𝑊𝐷𝑖𝑡 is the long-lived asset write down in year t, scaled by market value of equity at the beginning of the year, and 𝐺𝑊𝑖𝑡 is the impairment in year t, scaled by market value of equity at the beginning of the year. The remaining variables were defined previously. Following Banker et al. (2017), missing values of 𝑊𝐷𝑖𝑡 and 𝐺𝑊𝑖𝑡 are replaced with 0. As write-downs are coded as negative numbers, the expected coefficient signs for 𝑊𝐷𝑖𝑡 and 𝐺𝑊𝑖𝑡 are positive (similar to the models with earnings as the dependent variable).

The descriptive statistics and correlation coefficients for 𝑊𝐷𝑖𝑡 and 𝐺𝑊𝑖𝑡 are reported in Tables 4.3 and 4.4 respectively. Both tangible asset write downs and goodwill impairment are coded as negative given that they reduce earnings. Tangible asset write downs (𝑊𝐷𝑖𝑡) and goodwill impairment (𝐺𝑊𝑖𝑡) are scaled by the market value of equity at the beginning of the year. On

average, tangible asset write down is equal to −0.35% of lagged market value and average goodwill impairment is equal to −0.67% of lagged market value. Among firms reporting a non-

Documento similar