How-To Example: Dieselgate, incl. Data Preparation

With the VW Dieselgate event as the empirical setting, this page shows you how to fetch data in R, perform an event study using the EventStudy R-package, and do some basic plotting.

library(tidyquant)
library(dplyr)
library(readr)

Data Preparation

We use the package tidyquant to fetch the automotive stock data from Yahoo Finance. As we cannot get the full volume size from the required companies through Yahoo Finance API, we do not perform a volume event study in this vignette.

Let’s define the window from which we want to fetch the data of the German auto companies.

startDate <- "2014-05-01"
endDate <- "2015-12-31"

We focus us on the big five motor manufacturer in Germany, namely

  • VW Group
    • VW
    • Audi
    • Porsche
  • Daimler
  • BMW
# Firm Data
firmSymbols <- c("VOW.DE", "NSU.DE", "PAH3.DE", "BMW.DE", "DAI.DE")
firmNames <- c("VW preferred", "Audi", "Porsche Automobil Hld", "BMW", "Daimler")
firmSymbols %>% 
  tidyquant::tq_get(from = startDate, to = endDate) %>% 
  dplyr::mutate(date = format(date, "%d.%m.%Y")) -> firmData
knitr::kable(head(firmData), pad=0)
symbol date open high low close volume adjusted
VOW.DE 02.05.2014 194.05 194.05 188.00 188.55 67018 188.55
VOW.DE 05.05.2014 188.20 189.20 185.45 188.60 69295 188.60
VOW.DE 06.05.2014 189.05 189.90 184.80 186.00 49051 186.00
VOW.DE 07.05.2014 185.65 187.00 184.60 184.70 52484 184.70
VOW.DE 08.05.2014 185.05 189.00 185.05 188.95 51892 188.95
VOW.DE 09.05.2014 188.60 189.50 187.95 188.45 40368 188.45

As the reference market, we choose the DAX.

# Index Data
indexSymbol <- c("^GDAXI")
indexName <- c("DAX")
indexSymbol %>% 
  tidyquant::tq_get(from = startDate, to = endDate) %>% 
  dplyr::mutate(date = format(date, "%d.%m.%Y")) -> indexData
indexData$symbol <- "DAX"
knitr::kable(head(indexData), pad=0)
date open high low close volume adjusted symbol
02.05.2014 9611.79 9627.38 9533.30 9556.02 88062300 9556.02 DAX
05.05.2014 9536.38 9548.17 9407.09 9529.50 61911600 9529.50 DAX
06.05.2014 9570.25 9571.63 9440.47 9467.53 82062900 9467.53 DAX
07.05.2014 9418.50 9554.35 9410.08 9521.30 92732600 9521.30 DAX
08.05.2014 9547.27 9622.30 9487.57 9607.40 102022500 9607.40 DAX
09.05.2014 9591.32 9602.86 9558.11 9581.45 80084100 9581.45 DAX

Now, after we have fetched all the data, we prepare the data files for the API call, as described in the introductory vignette. We prepare in this step already the volume data for later purposes.

# Price files for firms and market
firmData %>% 
  dplyr::select(symbol, date, adjusted) %>% 
  readr::write_delim(path      = "02_firmDataPrice.csv", 
                     delim     = ";", 
                     col_names = F)

indexData %>% 
  dplyr::select(symbol, date, adjusted) %>% 
  readr::write_delim(path      = "03_marketDataPrice.csv", 
                     delim     = ";", 
                     col_names = F)

# Volume files for firms and market
firmData %>% 
  dplyr::select(symbol, date, volume) %>% 
  readr::write_delim(path      = "02_firmDataVolume.csv", 
                     delim     = ";", 
                     col_names = F)

indexData %>% 
  dplyr::select(symbol, date, volume) %>% 
  readr::write_delim(path      = "03_marketDataVolume.csv", 
                     delim     = ";", 
                     col_names = F)

Finally, we have to prepare the request file. The parameters for this Event Study are:

  • Estimation window: 250
  • Event window: -10 to 10
  • Event date: 18.09.2015

Details of the format can be found in the introductory vignette.

group <- c(rep("VW Group", 3), rep("Other", 2))
request <- cbind(c(1:5), firmSymbols, rep(indexName, 5), rep("18.09.2015", 5), group, rep(-10, 5), rep(10, 5), rep(-11, 5), rep(250, 5))
request %>% 
  as.data.frame() %>% 
  readr::write_delim("01_requestFile.csv", delim = ";", col_names = F)

Performing the Event Studies: Calculating Abnormal Returns, Volumes, and Volatilities

After the preparation steps, we are now able to start the calculations. We use in all types of event studies the GARCH(1, 1) model. Please consider in your event studies that fitting this model is computationally expensive and no near-real-time response from the API can be expected.

 

Calculating abnormal returns

key <- "573e58c665fcc08cc6e5a660beaad0cb"

library(EventStudy)
est <- EventStudyAPI$new()
est$authentication(apiKey = key)
## [1] TRUE
# get & set parameters for abnormal return Event Study
# we use a garch model and csv as return
# Attention: fitting a GARCH(1, 1) model is compute intensive
esaParams <- EventStudy::ARCApplicationInput$new()
esaParams$setResultFileType("csv")
esaParams$setBenchmarkModel("garch")

dataFiles <- c("request_file" = "01_requestFile.csv",
               "firm_data"    = "02_firmDataPrice.csv",
               "market_data"  = "03_marketDataPrice.csv")

# check data files, you can do it also in our R6 class
EventStudy::checkFiles(dataFiles)
## Checking request_fileChecking firm_dataChecking market_data
# now let us perform the Event Study
arEventStudy <- est$performEventStudy(estParams     = esaParams, 
                                      dataFiles     = dataFiles, 
                                      downloadFiles = T)
## [1] "Check batch process: Step 0"

Now, you can use the downloaded csv (or your preferred data format) files in your analysis. During the creation of the arEventStudy object we merge information from the request file, and the result files.

knitr::kable(head(arEventStudy$arResults))
Event ID eventTime ar tValue Firm Reference Market Estimation Window Length Group
1 -10 0.0032 0.3137 VOW.DE DAX 250 NA
2 -10 0.0253 1.8201 NSU.DE DAX 250 VW Group
3 -10 0.0052 0.5652 PAH3.DE DAX 250 VW Group
4 -10 0.0085 0.9551 BMW.DE DAX 250 Other
5 -10 0.0101 1.4429 DAI.DE DAX 250 Other
1 -9 0.0006 0.0588 VOW.DE DAX 250 NA

The averaged abnormal return (aar) data.frame has the following shape:

knitr::kable(head(arEventStudy$aarResults))
  level eventTime aar N Pos stat1 stat2 stat3 stat4 stat5 stat6 stat7 stat8 stat9 stat10
1 VW Group -10 0.0112 3 3 1.5247 1.6955 1.5871 1.9441 1.2859 1.4944 1.5225 1.9401 1.7928 2.7133
14 Other -10 0.0093 2 2 1.6669 1.3264 11.6250 4.9509 1.5658 2.0021 1.3952 3.1356 2.4105 Inf
27 VW Group -9 -0.0015 3 1 -0.2372 -0.6144 -1.4042 -1.3290 -0.5438 -1.3223 -0.2369 -1.3262 -1.5857 -0.6998
40 Other -9 0.0024 2 2 0.4740 1.3264 1.2973 1.2505 0.5367 1.0826 0.3967 0.7920 1.3077 NA
53 VW Group -8 0.0035 3 2 0.6652 0.5405 0.9476 1.1537 0.5618 0.7711 0.6643 1.1513 0.9243 0.7847
66 Other -8 0.0135 2 2 2.2489 1.3264 1.9286 2.3441 1.6954 1.7402 1.8823 1.4846 2.0982 Inf

You can find the statistic naming in arEventStudy$aarStatistics.

 

Calculating abnormal volatilities

est <- EventStudyAPI$new()
est$authentication(apiKey = key)
## [1] TRUE
# get & set parameters for abnormal return Event Study
esaParams <- EventStudy::AVyCApplicationInput$new()
esaParams$setResultFileType("csv")
 
avycEventStudy <- est$performEventStudy(estParams    = esaParams, 
                                       dataFiles     = dataFiles,
                                       downloadFiles = T)
## [1] "Check batch process: Step 0"

The prepared data.frames in avycEventStudy have a similar shape as for the abnormal return Event Study.

 

Calculating abnormal volumes

This will be added in a later stage.

# est <- EventStudyAPI$new()
# est$authentication(apiKey = key)
# 
# # get & set parameters for abnormal return Event Study
# esaParams <- EventStudy::AVCApplicationInput$new()
# esaParams$setResultFileType("csv")
# 
# avEventStudy <- est$performEventStudy(estParams = esaParams,
#                       dataFiles = c("request_file" = "01_requestFile.csv",
#                                     "firm_data"    = "02_firmDataVolume.csv",
#                                     "market_data"  = "03_marketDataVolume.csv"))