The effect of corporate social responsibility practices on real earnings management: evidence from a European ESG data

Abstract

The new trend in the ranking of companies, not only in financial terms but also according to their commitment to ethics and corporate social responsibility (CSR), is explained by the importance of responsible investment and responsible governance criteria in the decision making of the shareholders and other investors. These new business evaluation criteria have attracted the attention of several stakeholders, such as investors, financial analysts, researchers, and also specialized media, that require quality information based on this social and ethical approach. The aim of this article is to examine the link between corporate social responsibility practices and the level of real earnings management (REM) of a sample firms belonging ESG index from five European countries. The variables related to the ethical behavior of companies have a statistically significant and negative relationship with the level of real earnings management. Indeed, the more important the socially responsible and ethical practices are, the less the company engages in an aggressive REM strategy. Thus, the integration of new dimensions in the explanation and determination of the REM is under-explored. The explanation of the quality of the results by ethical or social variables makes it possible to overcome the criticisms addressed to the contractual approaches of companies.

Appendices

Appendix: Definition and measurement of REM index

Proxy 1: Abnormal cash flow from operations

In order to estimate the abnormal CFO, first the normal CFO is estimated. This is expressed as a linear function of sales (S_t) and the change in sales (ΔS_t). With this equation, the coefficients are estimated. All the variables in the model are scaled by lagged total assets (A_t−1) so that heteroskedasticity is reduced

$${\text{CFO}}_{t} /A_{{t - 1}} = \alpha _{0} + \alpha _{1} \left( {1/A_{{t - 1}} } \right) + \beta _{1} \left( {S_{t} /A_{{t - 1}} } \right) + \beta _{2} \left( {\Delta S_{t} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(3)

where CFO_t = cash flow from operations in period t; A_t−1 = lagged total assets; ΔS_t = S_t − S_t−1, change in sales; α₀, α₁, β₁ and β₂ = parameters; ε = error term.

By using the estimated coefficients from Eq. (3), the normal cash flow from operations (NCFO_t) is calculated

$${\text{NCFO}}_{t} /A_{{t - 1}} = a_{0} + a_{1} \left( {1/A_{{t - 1}} } \right) + b_{1} \left( {S_{t} /A_{{t - 1}} } \right) + b_{2} \left( {\Delta S_{t} /A_{{t - 1}} } \right)$$

(4)

where NCFO_t = normal cash flow from operations in period t; a₀, a_1, b₁ and b₂ = estimated coefficients of the parameters.

After calculating the NCFO_t, the abnormal CFO (ABN_CFO_t) is calculated. The ABN_CFO_t is the actual CFO minus the NCFO_t that was calculated with Eq. (4)

$${\text{ABN}}\_{\text{CFO}}_{t} = {\text{CFO}}_{t} /A_{{t - 1}} - {\text{NCFO}}_{t} /A_{{t - 1}}$$

(5)

where ABN_CFO_t = abnormal cash flow from operations in period t.

Proxy 2: Abnormal production costs

The steps that were taken to calculate the abnormal CFO are performed for the abnormal production costs. First, the definition of production costs is provided. Roychowdhury (2006) defines production costs (PROD_t)as the sum of costs of goods sold (COGS_t) and the change in inventory (ΔINV_t). (PROD_t = COGS_t + ΔINV_t). Since the PROD_t consists of COGS_t and ΔINV_t, Roychowdhury (2006) estimates these as following:

$${\text{COGS}}_{t} /A_{{t - 1}} = \alpha _{0} + \alpha _{1} \left( {1/A_{{t - 1}} } \right) + \beta _{1} \left( {S_{t} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(6)

where COGS_t = costs of goods sold in period t

$$\Delta {\text{INV}}_{t} /A_{{t - 1}} = \alpha _{0} + \alpha _{1} \left( {1/A_{{t - 1}} } \right) + \beta _{1} \left( {\Delta S_{t} /A_{{t - 1}} } \right) + \beta _{2} \left( {\Delta S_{{t - 1}} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(7)

where ΔINV_t = change in inventory.

Then, the normal level of COGS_t and ΔINV_t is calculated by using the estimated coefficients from Eqs. (6) and (7):

$${\text{COGS}}_{t} /A_{{t - 1}} = a_{0} + a_{1} \left( {1/A_{{t - 1}} } \right) + b_{1} \left( {S_{t} /A_{{t - 1}} } \right)$$

(8)

$$\Delta {\text{INV}}_{t} /A_{{t - 1}} = a_{0} + a_{1} \left( {1/A_{{t - 1}} } \right) + b_{1} \left( {\Delta S_{t} /A_{{t - 1}} } \right) + b_{2} \left( {\Delta S_{{t - 1}} /A_{{t - 1}} } \right)$$

(9)

The production costs are estimated by the sum of Eqs. (6) and (7) and are modeled as following:

$${\text{PROD}}_{t} /A_{{t - 1}} = \alpha _{0} + \alpha _{1} \left( {1/A_{{t - 1}} } \right) + \beta _{1} \left( {S_{t} /A_{{t - 1}} } \right) + \beta _{2} \left( {\Delta S_{t} /A_{{t - 1}} } \right) + \beta _{3} \left( {\Delta S_{{t - 1}} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(10)

where PROD_t = production costs in period t.

Consequently, the normal level PROD_t is calculated by using the estimated coefficients from Eq. (10)

$${\text{NPROD}}_{t} /A_{{t - 1}} = a_{0} + a_{1} \left( {1/A_{{t - 1}} } \right) + b_{1} \left( {S_{t} /A_{{t - 1}} } \right) + b_{2} \left( {\Delta S_{t} /A_{{t - 1}} } \right) + b_{3} \left( {\Delta S_{{t - 1}} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(11)

where NPROD_t = the normal production costs in period t.

Lastly, the abnormal production costs (ABN_PROD_t) are estimated as the actual PROD_t, minus the NPROD_t

$${\text{ABN}}\_{\text{PROD}}_{t} = {\text{PROD}}_{t} /A_{{t - 1}} - {\text{NPROD}}_{t} /A_{{t - 1}}$$

(12)

where ABN_PROD_t = abnormal production costs.

Proxy 3: Abnormal discretionary expense

The last proxy for the use of real earnings management is the abnormal discretionary expense. Just as the equations presented before, the discretionary expense is modeled as a linear function of sales

$${\text{DISEXP}}_{t} /A_{{t - 1}} = \alpha _{0} + \alpha _{1} \left( {1/A_{{t - 1}} } \right) + \beta _{1} \left( {S_{{t - 1}} /A_{{t - 1}} } \right) + \varepsilon _{t}$$

(13)

where DISEXP_t = the discretionary expense in period t.

Consequently, the estimated coefficients in Eq. (13) are used in Eq. (14) to estimate the normal discretionary expense (NDISEXP_t).

$${\text{NDISEXP}}_{t} /A_{{t - 1}} = a_{0} + a_{1} \left( {1/A_{{t - 1}} } \right) + b_{1} \left( {S_{{t - 1}} /A_{{t - 1}} } \right)$$

(14)

Finally, the abnormal discretionary expense (ABN_EXP_t) is measured as the difference between the actual and the normal discretionary expense

$${\text{ABN}}\_{\text{DISEXP}}_{t} = {\text{DISEXP}}_{t} /A_{{t - 1}} - {\text{NDISEXP}}_{t} /A_{{t - 1}}$$

(15)

where ABN_DISEXP_t = abnormal discretionary expenses.

In order to have a comprehensive metric and capture the total effect of real earnings management, we aggregate the three measures of real manipulation activities into one proxy, REMI, by taking their sum as follows (Cohen et al. 2008; Cohen and Zarowin 2010; Zang 2012): REMI = (− 1)*AbnCFO + AbnPR + AbnDE*(− 1).