Are corporate social responsibility and advertising complements or substitutes in producing firm reputation?

ABSTRACT

A firm’s reputation is one of the critical drivers of success, and two of the key levers firms use to influence their perceived reputation are corporate social responsibility (CSR) and advertising. The relationship between CSR and advertising is important because whether they are complements or substitutes has different implications for how firms use these activities. Using a unique panel dataset of US-listed companies between 2005 and 2014, we estimate flexible production functions to identify whether CSR and advertising act as complements or substitutes in the production of firm reputation. A secondary motivation of this paper is to examine whether the use of different stakeholder ratings of firm reputation matters. We find evidence consistent with advertising and CSR being substitutes toward the production of firm reputation. Our results also show that advertising, own-firm CSR activities, and industry-level CSR spillovers contribute positively to firm reputation. Lastly, we find that the effects of CSR and advertising vary across the stakeholder groups (general public, business executives, or CSR experts) used in the analysis.

KEYWORDS:

• Advertising
• social responsibility
• reputation
• production functions
• elasticities

JEL CLASSIFICATION:

• D24
• D21
• M14

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ For example, in a recent survey of US consumers, 77 per cent of consumers state that it is important for a company to be socially responsible and 70 per cent are willing to pay more for a $100 product from a responsible company (Landor Associates 2010).

² These data sources were merged using unique firm identifiers such as the International Securities Identification Number (ISIN) number, CUSIP code, and/or the firm name. Information on accessing the data used in this study is provided in the appendix.

³ For a more in-depth review of the conceptual and empirical issues regarding reputation research, please see Lange, Lee, and Dai (2011).

⁴ Exceptions include Fombrun and Shanley (1990) who use the amount a firm donates to charity as its CSR measure using data from the Taft database; and Stanwick and Stanwick (1998) who use data on environmental performance from the toxic release inventory in the United States.

⁵ Ideally, we would collect data on the expenditures each firm spends on CSR activities but this data is not available and thus we use third-party ratings of CSR.

⁶ As a robustness check, we also include corporate governance in the CSR measure and re-estimate the models. These results are presented in Table A-3 in the online appendix and the results are similar to the findings using only environmental and social performance metrics.

⁷ In a robustness check, we also include the corporate governance and product categories in the CSR and CSI indices and the results are qualitatively similar to those presented in the paper. These results are presented in Table A-4 of the online appendix.

⁸ A notable exception is Barrage, Chyn, and Hastings (2014).

⁹ Consistent with the prior literature, we set advertising expenditures to zero if they are missing in the database (Fee, Hadlock, and Pierce 2009; Servaes and Tamayo 2013).

¹⁰ For example, one spot ad during prime time television is quite different from one internet advertisement. Using total expenditures on all media sources is more appropriate because the quality of advertisements such as the number of customers reached is reflected in its cost.

¹¹ The specific industry categories are Mining & Construction (SIC codes: 1000–1799), Food, Textiles, & Apparel (2000–2399) Paper & Publishing (2400–2799), Chemicals & Pharmaceuticals (2800–2899), Refining, Rubber, Plastic (2900–3199), Heavy Manufacturing (3200–3569), Computers & Precision Products (3570–3699), Auto & Aerospace (3700–3799), Transportation Services (4000–4789), Telephone & Utilities (4800–4991), Wholesale & Retail (5000–5999), Bank & Financial Services (6000–6799), Other Services (7000–8999), and Non-classifiable Establishments (9000–9999).

¹² The timing of the RepTrak survey also alleviates concerns about omitted variable bias from not including contemporaneous advertising and CSR. The RepTrak survey is administered in January and February of each year and the results are generally released in April. Consequently, the advertising and CSR activities during the year of the survey release should only have a minimal effect on the reputation measures, if at all.

¹³ While the FE model is preferred due to its ability to control for unobserved firm-specific confounders that are time-invariant, we present both sets of results because the inter-temporal variation of firm reputation is limited. For example, the ratio of the variance in reputation due to the fixed-effects to the total variance in reputation (i.e. intra-class correlation coefficient) is around 0.8 in the various model specifications.

¹⁴ Note that we do not include an interaction term between advertising and industry-level CSR to reduce the number of parameters to be estimated. Ignoring the interaction terms implicitly implies a unit elasticity between these inputs.

¹⁵ For example, input $i$ may be more easily substitutable for input $j$ and thus ${\sigma }_{ij}^M>{\sigma }_{ji}^M$. The Allen elasticity of substitution measure imposes that ${\sigma }_{ij}={\sigma }_{ji}$.

¹⁶ In the case of a semi-translog production function, the first and second partials of the production function $y=f(x)$ are $f_i=\frac{dy}{d{x}_i}={ϵ}_i\frac{q}{x_i}$, $f_{ij}=\frac{y}{x_i{x}_j}({\beta }_{ij}+{ϵ}_i{ϵ}_j)$, and $f_{ii}=\frac{y}{x_i^2}({\beta }_{ii}+{ϵ}_i({ϵ}_i-1))$.

¹⁷ We also conducted the analysis not including the first two years of the general public reputation data with a low number of observations (37 in 2006 and 71 in 2007) as a robustness check. These results are provided in Table A-3 of the online appendix.

¹⁸ Table A-1 and Table A-2 in the online appendix present the comparable set of models using EXECUTIVE and EXPERT as the dependent reputation variables.

¹⁹ The ASSET 4 CSR variable measures relative performance and thus in the absence of any actual spillover effects, we would expect the industry-level CSR coefficient to be negative by construction. The finding that the coefficient is positive reinforces the idea that the spillover benefits to reputation are positive.

²⁰ We also estimate a model using the EXECUTIVE reputation variable, COMPUSTAT advertising data, and the KLD CSR/CSI variables for comparison with the existing literature. While not presented in this paper, the output and substitution elasticities for this model are statistically insignificant using the fixed effects specification, which corroborates the findings in Table 5. These results are presented in Table A-4.