29.1 Other Issues
29.1.1 Event Studies in marketing
(Skiera, Bayer, and Schöler 2017) What should be the dependent variable in marketing-related event studies?
Based on valuation theory, Shareholder value = the value of the operating business + non-operating asset - debt (Schulze, Skiera, and Wiesel 2012)
- Many marketing events only affect the operating business value, but not non-operating assets and debt
Ignoring the differences in firm-specific leverage effects has dual effects:
inflates the impact of observation pertaining to firms with large debt
deflates those pertaining to firms with large non-operating asset.
It’s recommended that marketing papers should report both \(CAR^{OB}\) and \(CAR^{SHV}\) and argue for whichever one more appropriate.
Up until this paper, only two previous event studies control for financial structure: (Gielens et al. 2008) (Chaney, Devinney, and Winer 1991)
Definitions:
Cumulative abnormal percentage return on shareholder value (\(CAR^{SHV}\))
- Shareholder value refers to a firm’s market capitalization = share price x # of shares.
Cumulative abnormal percentage return on the value of the operating business (\(CAR^{OB}\))
\(CAR^{OB} = CAR^{SHV}/\text{leverage effect}_{before}\)
Leverage effect = Operating business value / Shareholder value (LE describes how a 1% change in operating business translates into a percentage change in shareholder value).
Value of operating business = shareholder value - non-operating assets + debt
Leverage effect \(\neq\) leverage ratio, where leverage ratio is debt / firm size
debt = long-term + short-term debt; long-term debt
firm size = book value of equity; market cap; total assets; debt + equity
Operating assets are those used by firm in their core business operations (e..g, property, plant, equipment, natural resources, intangible asset)
Non–operating assets (redundant assets), do not play a role in a firm’s operations, but still generate some form of return (e.g., excess cash , marketable securities - commercial papers, market instruments)
Marketing events usually influence the value of a firm’s operating assets (more specifically intangible assets). Then, changes in the value of the operating business can impact shareholder value.
Three rare instances where marketing events can affect non-operating assets and debt
(G. C. Hall, Hutchinson, and Michaelas 2004): excess pre-orderings can influence short-term debt
(Berger, Ofek, and Yermack 1997) Firing CMO increase debt as the manager’s tenure is negatively associated with the firm’s debt
(Bhaduri 2002) production of unique products.
A marketing-related event can either influence
value components of a firm’s value (= firm’s operating business, non-operating assets and its debt)
only the operating business.
Replication of the leverage effect
\[ \begin{aligned} \text{leverage effect} &= \frac{\text{operating business}}{\text{shareholder value}} \\ &= \frac{\text{(shareholder value - non-operating assets + debt)}}{\text{shareholder value}} \\ &= \frac{prcc_f \times csho - ivst + dd1 + dltt + pstk}{prcc_f \times csho} \end{aligned} \]
Compustat Data Item
| Label | Variable |
|---|---|
prcc_f |
Share price |
csho |
Common shares outstanding |
ivst |
short-term investments (Non-operating assets) |
dd1 |
long-term debt due in one year |
dltt |
long-term debt |
pstk |
preferred stock |
Since WRDS no longer maintains the S&P 500 list as of the time of this writing, I can’t replicate the list used in (Skiera, Bayer, and Schöler 2017) paper.
library(tidyverse)
df_leverage_effect <- read.csv("data/leverage_effect.csv.gz") %>%
# get active firms only
filter(costat == "A") %>%
# drop missing values
drop_na() %>%
# create the leverage effect variable
mutate(le = (prcc_f * csho - ivst + dd1 + dltt + pstk)/ (prcc_f * csho)) %>%
# get shareholder value
mutate(shv = prcc_f * csho) %>%
# remove Infinity value for leverage effect (i.e., shareholder value = 0)
filter_all(all_vars(!is.infinite(.))) %>%
# positive values only
filter_all(all_vars(. > 0)) %>%
# get the within coefficient of variation
group_by(gvkey) %>%
mutate(within_var_mean_le = mean(le),
within_var_sd_le = sd(le)) %>%
ungroup()
# get the mean and standard deviation
mean(df_leverage_effect$le)
#> [1] 150.1087
max(df_leverage_effect$le)
#> [1] 183629.6
hist(df_leverage_effect$le)
# coefficient of variation
sd(df_leverage_effect$le) / mean(df_leverage_effect$le) * 100
#> [1] 2749.084
# Within-firm variation (similar to fig 3a)
df_leverage_effect %>%
group_by(gvkey) %>%
slice(1) %>%
ungroup() %>%
dplyr::select(within_var_mean_le, within_var_sd_le) %>%
dplyr::mutate(cv = within_var_sd_le/ within_var_mean_le) %>%
dplyr::select(cv) %>%
pull() %>%
hist()