E-commerce: Drivers of Success and Failure PDF Free Download

Name: E-commerce: Drivers of Success and Failure PDF
Author: im_daniel

1 / 58

0 views•58 pages

E-commerce: Drivers of Success and Failure PDF Free Download

E-commerce: Drivers of Success and Failure PDF free Download. Think more deeply and widely.

E-commerce: Drivers of

Success and Failure

A Master Thesis

presented to

the Department of Economics

Johannes Kepler University of Linz

To confer the academic Degree of

Master of Science (MSc)

in the Master’s Program

Management and Applied Economics

Valeriya Azarova (1255395)

Prof. Dr. Franz Hackl, thesis supervisor

Linz, June 2015

SWORN DECLARATION

I hereby declare under oath that the submitted Master's degree thesis has been written

solely by me without any third-party assistance, information other than provided sources

or aids have not been used and those used have been fully documented. Sources for

literal, paraphrased and cited quotes have been accurately credited.

The submitted document here present is identical to the electronically submitted text

document

Linz,

ACKNOWLEDGMENTS

I would like to thank my supervisor, Prof. Dr. Franz Hackl, for the enormous patience,

guidance, encouragement and advice he has provided throughout my time as his

student. There are no words to express how much Prof. Dr. Franz Hackl has taught me.

I have been extremely lucky to have a supervisor who cared so much about my work,

and who responded to my questions and queries so promptly. His immense knowledge,

patience and selfless dedication to research have been my motivation throughout my

Master Thesis work. His forensic scrutiny of my economic writing has been invaluable.

He has always found the time to propose consistently excellent improvements and help

me to move further with my critical thinking skills. I owe a great debt of gratitude to Prof.

Dr. Hackl.

I would also like to thank all the members of staff of Economic department at Johannes

Kepler University of Linz. In particular I would like to thank Ms. Mag. Jasmin Voigt for her

boundless friendship, kindness and support, but especially for helping me to cope with

all the organizational formalities of Master Thesis writing and submission.

I must also express my deepest gratitude to my family, my husband, son and parents for

their continued support, encouragement and patience, and for sharing with me all of the

ups and downs of my research.

TABLE OF CONTENT

LIST OF FIGURES 5

LIST OF TABLES 6

1. INTRODUCTION 7

2. LITERATURE REVIEW 9

3. HYPOTHESES DEVELOPMENT 12

3.1 Business 12

3.2 Site 13

3.3 Products 14

3.4 Service 16

3.5 Sales 17

4. DATA 18

4.1 Database 18

4.2 Dependent variable 19

4.3 Independent variables 21

5. ECONOMETRIC METHODOLOGY 25

6. ESTIMATION RESULTS 28

6.1 OLS Model: First Insight of the data 28

6.2 Kaplan-Meier: Group comparisons 29

6.3 Competing risks: Main research method and results 32

6.4 Robustness tests 34

7. CONCLUSIONS AND LIMITATIONS 39

REFERENCES 42

APPENDIX 45

DESCRIPTIVE STATISTICS 56

LIST OF FIGURES

Figure 1. Austrian e-tailers (from 2006 to 2012). ............................................................ 20

Figure 2. Duration of e-tailers in Austria (from 2006 to 2012). ....................................... 20

Figure 3. Kaplan-Meier: single- vs. multi-channel e-tailers. ............................................ 30

Figure 4. Kaplan-Meier: Eurolabel vs. no Eurolabel e-tailers. ........................................ 30

Figure 5. Kaplan-Meier: small vs. big e-tailers (by number of products offered). ........... 31

LIST OF TABLES

Table 1. Determinants of e-commerce firms’ lifespan in Austria. ................................... 45

Table 2. Drivers of e-commerce success and failure. .................................................... 47

Table 3. Drivers of e-commerce success and failure. Small vs. big firms. ..................... 48

Table 4. Drivers of e-commerce success and failure. Low price vs. high price firms. .... 49

Table 5. Drivers of e-commerce success and failure. Old vs. young firms (by birthdate).

....................................................................................................................................... 51

Table 6. Drivers of e-commerce success and failure. Crisis (by timestamp). ................. 53

Table 7. Drivers of e-commerce success and failure. Competing risks vs. Cox

proportional hazards. ..................................................................................................... 54

Table 8. Drivers of e-commerce success and failure. Brand concentration. .................. 55

1. INTRODUCTION

Since the beginning of the 21st century e-commerce has been rapidly growing all over

the world. It has turned into a new and powerful sales channel and changed the way

companies do business, the way they operate.

E-commerce has also modified consumers’ expectations about speed, price and quality

of service and products. Competition seems to be tougher, when it takes just one click to

switch to competitors and only several seconds to compare products of different

suppliers with help of online price search engines (like pricegrabber.com or google

shopping).

Some industries suffered from e-commerce like music CD’s or tourist agencies, while

others could profit from online retail as it has opened new horizons and provided new

management and marketing tools like Google Statistics or Google Ads. Managers have

received a possibility to monitor and moderate online 24/7 their advertising budgets,

prices, products and services.

Online retail is one of the fastest growing industries with Worldwide B2C sales in 2013 -

$ 1.25 trillion and international growth of 20.1 % in 2014 (eMarketer). Online retail is

especially attractive as a start-up option – ostensibly, it requires less human, time and

financial resources compared to traditional brick-and-mortar business and promises high

and quick gain.

Unfortunately, there is no gain without any pain: e-commerce is not only attractive, but

also risky - up to 90% do not make it through the first five years. In order to understand

why so many firms fail, it is necessary to investigate various drivers of success and

failure of online business.

Firms’ survival has already been examined with different methods and from different

perspectives, depending on the discipline: financial researchers use capital market

characteristics, financial ratios and define firm failure as bankruptcy or failure of IPO

(Aragawal et al, 2002); economists are interested in various firm, industry and market

related drivers of success; recent Information System researchers apply Bayesian

statistics to investigate firms’ duration (Kauffman and Wang, 2003, Banerjee et al.,

2007). But not much research is devoted specifically to online retail survival.

Audretsch (1991) found that survival rates vary considerably across industries. Yet

specific characteristics of e-commerce industry have not been analyzed. None of

previous works considered such online specific factors as traffic, clients’ feedback and

trust, products price and price change frequency, average price of clicked products, time

spent on the site, shipping costs, shop platform or availability of app/mobile version. All

these factors do not exist in traditional brick-and-mortar retail, but might be driving

survival in e-commerce.

One of the first questions you ask yourself when launching an online shop is what about

the traffic, offer and updates – where do I get the clients from? How much am I ready to

pay for a click of potential customer to Google Advertisement or price search engines?

How often should I change the price and how many products offer? Should I stay highly

- specialized or more diversified? Should I offer more branded products or better stick to

lower prices? Most of those who launch online shops fail to give the right answers and

as a result fail in general.

So the goal of my thesis is to help finding answers to these questions by identifying

possible drivers of e-commerce firms’ success and failure, including business model,

products, service, clients, site and online specific characteristics. I plan to test several

hypotheses, helping to predict survival rates of online retailers using data from an

Austrian price search engine.

Based on data on 1096 e-tailers, observed for the period of six years, from 2006 to

2012, I will estimate a hazard model for firm survival in e-commerce with time varying

covariates. Hence, this thesis fills the gap in the existing literature on e-commerce and

provides a broader review of the factors that determine survival in online retail industry.

2. LITERATURE REVIEW

E-commerce represents a novel and trendy research field. Although there is extensive

business literature devoted to online retail and among all to the factors determining

success in this industry, still scientific empirical research on the issue is limited

(Loiacono et al., 2002, Wolfinbarger and Gilly, 2003, Parasuraman et al., 2005).

First efforts to form theoretical framework about e-commerce and its drivers of success

have been made in 2002 by Varadarajan and Yadav. In their research they try to

integrate aspects of the Internet into marketing strategy and suggest that specific

competitive strategy (order of entry), industry, firm, product and buyer characteristics

can be the drivers of online firms’ survival.

Varadarajan and Yadav state that online retail differs a lot from traditional brick-and-

mortar business: there is no physical contact between seller and buyer, no possibility to

try on or touch products before a buy is committed. Trust between customer and

supplier can be the key element of survival on e-commerce market. The authors suggest

that it is important for online retail survival whether digital goods or physical products,

search or experience goods, existing or new-to-the-world products are offered and

whether customization of products is possible or not. As for specific firm characteristics,

survival rates in online retail should differ depending on whether a firm is a pure online

retailer or has also offline retail expertise. Firms with relatively larger number of alliances

should, according to Varadarajan and Yadav, be less vulnerable.

First empirical studies of e-tailers survival find mixed results. They do not divide firms on

pure online shops and shops with both offline and online channels, do not incorporate e-

commerce specific factors, focus mainly on market, industry or financial determinants

and limit the data to public e-tailers (Loiacono et al., 2002, Wolfinbarger and Gilly, 2003,

Parasuraman et al., 2005).

In 2007 Nikolaeva publishes an empirical research based on theoretical inputs by

Mahajan and Varadarajan. She observes a sample of 460 dotcom e-tailers and builds

an explanatory model to predict survival rates, using complementary log-log regression

and Cox proportional hazards regression. She divides e-tailers on single- and multi-

channel and finds that survival rates for multi-channel e-firms are constantly higher.

Looking at the industry structure, Nikolaeva claims that exit rates increase with

increases in competitive density. The growth of the market both in terms of firms and

sales increases the hazard of exit, which can be explained by intensified competition,

lower margins, and higher marketing expenses. Age effects are interpreted as

accumulation of organizational knowledge. Publicly traded companies seem to be more

successful during their early developmental stage, because they are in a better position

to accumulate and interpret environmental knowledge. But this advantage is not stable

and dissipates with time, indicating that internal company knowledge and experience

become more important as companies age. Turning to products’ characteristics, e-tailers

selling digital products do not show constant higher survival rates. Another classification

of product categories – search vs. experience characteristics – does not yield conclusive

results either (Nikolaeva, 2007).

A different methodological approach is suggested by Kauffman and Wang (Kauffman

and Wang, 2007). They state that the influence of factors on e-tailers survival may vary

over time. So they use Bayesian dynamic models to test whether the share of new

Internet retailers IPOs among existing public e-commerce firms, debt ratio, executive

office salary, financial capital, product type and firm size have different impact on e-

commerce firms’ survival at different stages of their lifetimes. According to their results

firms that sell digital products have inverted U hazards – they experience lower hazard

rates in the first years after their IPOs, in the 3d year things change and firms with

physical goods do better, but after the 3d year the likelihood of failure starts to decrease

again for firms offering digital products. Growing executive office salary has been found

to be associated with lower hazard rate. The authors find that more financial capital was

associated with a lower likelihood of failure at earlier stages, but further on as firm

matured this association have diminished. No clear results have been found for debt

ratio impact on survival of sampled firms.

Summarizing current research frontier – hardly any of previous studies have included in

their analysis internet firm specific factors like traffic size and source, type of online

platform, number of clicks, clicks’ price, shipping costs or number of distinct visitors. If

some of those factors have been included, the sample was limited to public e-tailers

from US market, which cannot be representative for the whole e-commerce industry

even in US, without considering world e-commerce industry in general. European

market as well as small and medium, privately owned companies stay uncovered.

At the same time Europe is a big and fast growing market, with an average growth rate

of 19% in 2012. It may reflect a totally different pattern of online retail survival, due to

cultural, economic and geographic differences. Central Europe’s e-commerce market

was worth 75.9 billion euros in 2012, but it is expected to have grown to 93.3 billion

euros last year, which comes down to an increase of 23% in 2013. Among European

countries Austria is the second biggest online market after Germany. Austrians are also

the biggest spenders of the region with average expenditure of 2.085 euros in 2012 per

online shopper (Ecommerce News, 2014).

Much of the available research represents descriptive or qualitative analysis (Mahajan et

al., 2002, Shareef et al., 2006, Cosgun, 2010). So using the data on Austrian online

firms I hope to get an overview of survival pattern on European online market and

commit to filling the gap in the existing empirical research on e-commerce.

In the following chapters I will describe the hypotheses development, data, methods and

results of empirical research on e-commerce survival in Austria. I will discuss limitations

and provide some conclusions in the end.

3. HYPOTHESES DEVELOPMENT

Based on the literature described in the previous chapter, one can assume a certain

correlation between industry structure, firm, product, macro environment, buyer,

marketplace, financial characteristics, and online firms’ survival in the market.

As financial and industry characteristics have already been tested by previous

researchers (Kaufmann and Wang, 2002, Banerjee et al., 2006, Nikolaeva, 2007), and

the data set used for this work provides novel information on online specific

characteristics like number of clicks, shipping costs or distinct cookies, I suggest staying

focused on this kind of parameters, while formulating hypotheses. In some cases the

signs of the presumed impacts of shop characteristics on the failure of e-tailers can be

contradictory.

Based on the classification of success factors for survival of online shops offered by

Varadarajan and Yadav (2002), Shareef et al., (2006) and Nikolaeva (2007) and the

available data set, we can divide all the hypotheses on the following categories:

business, site, products, service and sales.

3.1BUSINESS

Different business models have different rates of survival depending on customer

segments, sales channels, customer relationships, revenue streams, key resources,

key activities, key partnerships, and cost structure (Business Survival Toolkit, 2014).

Some firms might have more expertise; some might be more flexible, and some more

resistant to changes. Analyzing sales channels (together with expertise) and e-

tailers’ niche, should help to identify whether online shops with different business

models have different chances on success.

For instance if a firm is a startup and has no offline experience (or is single-channel), it

might have lower chances on success than firms that have already managed to be

successful as traditional brick and mortar retailers (multi-channel).

Most of the previous researchers claim multi-channel firms have an advantage over

single channel e-tailers in terms of greater financial resources, stronger managerial skills

and established reputation. Some authors (Neslin et al, 2009) yet suggest that multi-

channel firms can also have lower survival rates due to a more complicated structure

and slower data integration, coordination of strategy, and allocation of available

resources among channels.

H1: Having offline expertise can have negative or positive impact on survival of e-tailers.

A firm can choose to be specialized in a certain niche of goods (narrow variety) or to

stay more general in terms of product lines (wide variety) variety. According to “niche

theory” (Hannan and Freeman, 1977) generalists should be outcompeted by specialists

under stable environmental circumstances, but can have higher chances in uncertain

and rapidly changing environment. Levins in his work “Theory of fitness in a

heterogeneous environment” (1962) suggests that also under unstable and rapidly

changing environments, generalists have lower chances to adapt and to survive than

specialists. Levins motivates his argument with higher time and energy needed to adapt

the structure and reallocate resources to match new circumstances (Levins, 1962).

H2: Highly specialized e-tailers can have higher or lower chances on success than

generalized e-tailers.

3.2 SITE

There are many books, articles and blogs devoted to “musts and must nots” of building a

profitable online shop. Of course there is no single formula of success, yet one thing is

clear - building a profit-maximizing, qualitative, functional (both for users and managers),

and trust-gaining site is a number one task for those who wish to succeed, as site is the

heart and face of e-tailer (Rembor, 2010).

Most of the small and medium e-tailers use ready solutions – universal platforms with

rather standard options and interface, which can save some money in the short term,

but can lead to higher expenses or even failure in the long term. If the business starts

growing fast, standard software might not be sufficient to cope with traffic and number of

orders. Custom made solutions can also provide a faster and easier way of uploading

and updating information, or offer more information on each client’s history and behavior

on the site, saving operational time and money.

A good interface is also important as it could help to create trust and keep potential

clients longer on the site. Font size and type, main color, speed of response, main

menu, easy and intuitively clear navigation - every aspect should be considered.

Investments in a good platform, nice-looking and functional interface, proper photos

quality, products’ description and the regularity of updates may not only increase

credibility of e-shop together with number of repeated visitors, but also help gaining

recognition among external organization providing seals of quality. The seals of quality,

in return, help to decrease uncertainty and gain trust among potential customers.

H3: External seals of quality should have a positive impact on e-commerce firms’

survival.

3.3 PRODUCTS

As in any other business in e-commerce it is not only important how to operate, but also

what to sell. Previous researchers claim that crucial about the products offered online is

whether they are search or experience goods (Kaufmann and Wang, 2007, Nikolaeva,

2007). They suggest that selling search or easy digitalized goods can provide some

advantages for e-tailers. However other firms’ characteristics like total number of offered

products, number of branded products, average price of offered products and price

changes, can also be driving online firms’ failure or survival.

The variety of products offered might be also an important factor of online shops’

success. A wider range of products increases the costs: more time and personal

resources needed to upload and update information on products and lager stock houses

(Randall and Ullrich, 2001).

H4: Number of products offered might have a negative impact on e-commerce survival.

Products offered by e-tailers may be branded or not. A brand might have a positive

impact on survival, because it is a signal, universal sign of trust, which may help to

decrease uncertainty for clients by promising a certain level of quality.

However, offering branded products can also cause a negative influence on survival due

to the higher prices for branded products compared with unbranded goods. Brand

consciousness is not something all the customers have; some customers could argue if

it is more or less the same telephone with same functions, made in China, why should I

pay more for a different logo? So they prefer buying unbranded products which should

decrease chances on success for e-tailers offering branded products only.

H5: Selling branded products might increase or decrease hazards of failure for e-tailers.

Setting competitive prices should be an important identification of e-commerce firms’

success. The shops’ average price of offered products can have a negative and positive

effect on success. Because of fast and easy price comparison in e-commerce, setting

price lower than the competitors can play an important role in success, as consumers

tend to buy at shops which offer lower prices for the same products. However setting

prices too low may simply not cover the costs and lead to failure. On the other hand

some customers might see better service or quality behind higher price.

H6: Average price of offered products might have negative or positive influence on

survival of e-tailers.

As online markets are dynamic, setting prices just once a month could not be enough.

Markets should be monitored and prices have to be re-adjusted according to the

behavior of competitors and the changing market conditions (e. g. reaction on wholesale

price changes, market introduction of substitutes). The frequency and height of price

changes might have both negative and positive effects. A shop that often changes

prices is expected to monitor constantly the market and keep its offers updated. A shop

which manages its duties of active price setting increases the possibility to survive;

however too frequent changes in prices might be a sign of panic and inability to set right

prices in time and evaluate the current situation on the market. A significant price hike

should be done when customers are ready to tolerate it, when they are loyal.

So choosing a proper strategy of price changes, taking into consideration seasonality,

customers’ and products’ specific characteristics should be an important driver of

survival on online market (Entrepreneur, 2014).

H7: The frequency and height of price changes might have negative or positive influence

on e-commerce firms’ survival.

3.4 SERVICE

If the business model works, site is attractive, credible and popular, products and prices

are balanced, there is one more thing left to turn a visitor into a loyal customer – service.

The key elements of e-tailers’ service are order and shipping procedures.

E-commerce has changed the way business is operated and also the expectations of

customers – they want prices to be fair, orders to be done easily and shipped fast. After

making an order one expects to receive the goods as fast as possible. It is part of the

philosophy of e-commerce to comfortably and fast make an order sitting at home,

without losing time on driving in traffic, searching for parking place, and to receive the

product at home within the shortest period possible.

And if something goes wrong there is almost always a forum on the site or a feedback

form, a popular blog or a price search engine, where one can share his or hers

experience of the ordering process on a specific web shop.

So, providing best service to every single client is necessary to keep up the good

reputation and trust.

H8: Positive feedbacks and reputation should decrease risks of failure in online market.

E-shops that want to survive should consider displaying the goods that are shortly

available. It would be disappointing for consumers to spend time searching for the

product and after finding and clicking on “buy” to receive a message that it’s “currently

not available”.

H9: Availability of products should have a positive influence on e-tailers’ survival.

Shipping costs can be another driver of success: if they are too high potential clients

might prefer to buy the product in the nearest shopping mall.

H10: Shipping costs should have negative impact on e-commerce failure.

3.5 SALES

No business can work without sales and clients. The more loyal clients there are the

higher are the sales and chances of success. E-commerce specific measures such as

distinct cookies can be used to track unique visitors and help to understand how

important the number of buyers is for the success of an online store.

H11: More distinct clients should increase e-tailers’ chances of survival.

Number of clicks and average prices of clicked products can serve as a certain proxy for

numbers and prices of products sold. The more clicks retailers get and the higher the

prices are the higher are the chances to survive.

H12 : Number of clicks and prices should have a positive effect on e-commerce success.

4. DATA

4.1 DATABASE

In order to determine drivers of e-commerce success and failure, I use the database of

Austrian e-tailers from http://geizhals.at. Geizhals.at is a popular price comparison

platform that collects information on products and prices from online retailers and

displays this information in one page in response to customer’s search of a certain

product. Potential buyers get an overview of each listed e-tailer’s price and can compare

offers also based on availability, shipping costs, payment options and service.

Geizhals.at has been founded in 1996 and has fast become one of the most frequently

used price-comparison services in Austria. Any retailer willing to survive on Austrian

online markets will most likely sign a contract with geizhals.at and pay several cents per

click on their products (pay-per-click)

in order to get constant targeted customer traffic.

This should mean that the data set used represents a full survey of Austrians e-

commerce at least in the main product categories covered by geizhals.at. Starting with

hardware comparison, geizhals.at now displays products in Software, Games, Phones,

Video/Cameras/TV, Audio/HiFi, Movies, Home & Garden, Health & Beauty and Sports &

Leisure. With most recently added categories – Sport & Leisure and Health & Beauty

geizhals.at is no more a price search engine specialized in technics, but a general price

comparison platform, covering most popular product markets except apparel.

The observations, taken into account, range between 23d of April 2006 and 09th of

September 2012. Monthly data is considered. 248 firms have already existed at the

beginning of the observation period. Excluding these firms that have been born on or

before the first day of observation will provide an inflow sample, which is a random

sample of those, who start an episode at a specific point in time.

E-tailers can also use pay-per-order option and pay a certain percentage of the price of ordered goods.

4.2 DEPENDENT VARIABLE

My dependent variable is the duration of an online firm together with its status at the end

of the follow-up period.

Status (event) is coded

 as 1, if the firm has no sign of presence online - no online shop and no more

listed on the price search engine;

 as 2, if the firm is no more listed on price search engine or stopped direct online

sales, but the business is still running. The e-tailers that have either catalogue of

products or services online, visit-card site with contacts of offline office or a

message on the site that online shop is being under reconstruction are attached

to this group of “question marks”. As I cannot track how long the catalogue or

message about site reconstruction is already hanging on the site – we suggest

there still is a possibility that the e-shop will continue working;

 as 3 if firm has an online shop, but is no more listed on the price search engine,

or stopped listing for a certain period of time and continued after the end of follow

up period This type of failure should be especially interesting for geizhals.at to

understand the logic of working with online price search engines, as this group

consists of successful firms that for some reasons have stopped their cooperation

with price search engine; and 0 otherwise.

The total number of e-tailers observed during the period is 1096 (see Figure 1): at the

end of the observation period 8% are still present online with catalogue, visit-card site,

contact address of the offline office or their online shops is under reconstruction, 12%

have online shops, but are no more listed on geizhals.at, and 38% do not have online

shops anymore. Only 42% of observed online retailers stay alive during the whole

observation period, without any change in the status.

While manually checking on Google information on all the retailers no longer listed on geizhals.at, I found

out that some of them were still active as e-tailers, and others have closed online shops, but kept their sites

with contacts of offline office or catalogue of products, or have their online shops under reconstruction.

In the main regressions status 2 and 3 are united in one category.

Duration is defined as the number of days passed from a firm’s first appearance on the

price search engine (birth) to either total failure of the online shop or end of being listed,

or the end of the observation period if the firm is still alive at that time.

12%

38%

42%

1096

0 200 400 600 800 1000 1200

"?"(event=2)

"Geizhals

death"(event=3)

Failed (event=1)

Alive

Total

Figure 1. Austrian e-tailers (from 2006 to 2012).

Figure 2. Duration of e-tailers in Austria (from 2006 to 2012).

The average duration of firms that stay alive during the whole follow-up period is 2220

days (6 years), while the average duration of firms that fail is just 1000 days (2.7 years).

The overall minimum duration is 11 days and the maximum is 2446 days (Figure 2). I will

use the duration in a survival analysis model to explain this variation. In the following I

will present different independent variables which might have impact on the lifetime of

firms. By analyzing the statistical impact of these variables, I will present empirical

evidence on the validity of the suggested hypotheses.

4.3 Independent variables

Business

To test the first hypothesis about dependency between different business models and e-

tailers survival, I will use the binary variable pick-up, which is equal to 1, if the retailer

offers clients a possibility to pick-up order by himself in his office or store, and 0

otherwise. As a lot of online shops develop their business out of a brick and mortar

store, the pick-up possibility might probably correlate with a prior offline experience in

the retail business. Furthermore this variable should be a good proxy to determine

whether a firm is a multi- or single-channel enterprise.

To test the second hypothesis about possible influences of specialization (deep

assortment) on survival rates of e-tailers, the variables products_hhi and clicks_hhi are

suggested. Products_hhi is a Herfindahl–Hirschman Index, measuring firms’ products

concentration in a certain sub-subcategory

. The maximal possible value is 10,000; the

closer to this value the more an e-tailer is specialized on a certain type of product.

Roughly 15% of all firms in the data set have a deep assortment and are highly

specialized, with products_hhi higher than 9000, for example druckertinteshop.at,

handy4u.at or media-consult.at.

Another measure of specialization which can be used to test the second hypothesis is

concentration of clicks based on sub-subcategories of products (clicks_hhi). As

Geizhals.at classifies products on categories, sub-categories and sub-subcategories, for example category

“Phone&Co”, subcategory “Mobile phones”, sub-subcategories “Chargers”, “Cases”, ”Mobile phones with

contract”.

expected, it is positively correlated with products_hhi (0.8266). So just one of the two

variables should be used in the regressions.

Products

A larger number of products offered might increase expenses on stock houses and may

demand for a more sophisticated functionality. I plan to use the variable offer to test

hypothesis 4. It captures the absolute number of products offered by an e-tailer at a

point in time. The minimum value is equal to 1, while the maximum (offered by

amazon.at) is 145,890 products.

Hypothesis 5, considering the impact of brand on survival, can be tested with the help of

two variables available in the dataset: branded_pct and brand_hhi. A brand, well known

to consumers, helps to decrease uncertainty. It signals a certain level of quality and

creates trust. On the other hand branded products are usually more expensive than

unbranded. Hence, customers should agree to pay a supplementary fee for the brand

name. Branded_pct is the overall percentage of branded products out of all offered by

an e-tailer at a certain point in time. Brand_hhi is a Herfindahl–Hirschman index,

indicating the concentration of branded products offered by an e-tailer. If this variable is

equal to 10,000 (maximal possible value) it means that e-tailer offers just one brand, the

lower this value is the more different brands are offered.

Following hypothesis 6, I check whether the average price of offered products has some

impact on e-commerce survival rates using variables price_avg or price_rel. The first

variable captures the average price of all offered products and is measured in euros,

while the second one is the average relative price of all products offered by e-tailer.

I will test whether hypothesis 7 about the dependency between e-commerce survival and

the frequency of price change holds in my dataset with #price_changes,

menu_cost_sync, menu_cost_height and menu_cost_time variables. #price_changes

captures the number of overall price changes divided by the number of products offered

in the period. Minimal value observed in the dataset is 0.0001 and the average is 0.057.

Menu_cost_sync is the indicator for the synchronicity of price changes, which accounts

for time and size of price changes and is highly correlated with #price_changes

(0.9515), so only one of the two variables will be included. Menu_cost_time shows the

average time in days between two price changes, the observed mean value for this

variable in the data set is 12 days, the variable ranges from 1 to 31 days.

Menu_cost_height is the average height of relative price represented by the ratio of

respective price changes and the original price. This variable varies from -1 to 19,574 (6

observations were excluded as outliers).

Site and services

Hypothesis 9 suggests a dependency between availability of products, customers’

satisfaction and e-commerce success. I suggest using variables immediately_available

which provides information on whether the product is available and shortly_available,

which shows that e-tailer has marked this product as available shortly (within 2-3 days).

Geizhals.at as well as other price search engines can sort firms by the price of offered

products including shipping costs. The option with lowest total price will be shown first

and should have higher chances for potential orders. Shipping costs are measured in

euros. The average shipping cost observed in the data set for delivery in Austria

(shipping_at) with payment in advance is 7.5 euros, though the maximum value is 94

euros. The average shipping cost with payment by cash (shipping_cash) on delivery is

11.6 euros. Other variables, like shipping_de, providing information on shipping to

Germany, have too many missing values and cannot be used to test hypothesis 10.

The dataset also provides information on products’ and etailers’ evaluations by

customers, which I’m planning to use in order to test hypothesis 8. Each registered

customer can evaluate retailers on geizhals.at in form of ratings from 1 (the best

possible) to 5 (the worst possible) together with textual comments on his experience.

The average value of 1.44 for the 12 months rating observed in the data set is quite

high. Users can also rate separately navigation, web-site performance, variety of

products, order procedures, shipping speed, price and quality, service provided,

payment options – the overall rating is formed from all these parameters.

The provision of this information at the geizhals.at site must have contributed

considerably to the price comparison platform’s popularity among users. Due to low

variation in ratings calculation of quintiles was used to build new binary variables –

rating1 (very good), rating2 (good), rating3 (average) and rating4 (bad - for retailers with

bad ratings 4 and 5).

However the ratings can be subjective and noisy (Hackl et al.,

2011).

A more objective way to rate e-tailers is suggested by organizations providing seals of

quality like Eurolabel. Eurolabel is a binary variable that captures whether an e-tailer has

the seal of quality or not. It is created by matching the list of Austrian e-tailers

possessing the seal from www.euro-label.com and the firm_name variable in the

dataset. I plan using Eurolabel to test hypothesis 3 claiming importance of external

quality seals in building credibility and trust.

Sales

Primary interest of e-tailers to be listed on geizhals.at is to get targeted traffic (someone

who looks at the price of a concrete product has a higher potential to commit an order

than someone who just by accident visits the site while surfing on the internet) and

increase sales.

I do not possess direct information on the number of clients or sales; however, I can use

distinct cookies among clicks of one e-tailer (cookies) together with prices of clicked

products to get some idea about demand and potential sales of e-tailers. Variables

cookies as an indicator of number of clients and also the average price of clicked

products (click_price_avg) are suggested to be used to test hypotheses 10 and 11.

97 e-shops have rating 4 and 23 e-shops have rating 5.

5. ECONOMETRIC METHODOLOGY

As outlined above the duration or the time measured in days an online firm manages to

survive will serve as the dependent variable in determining factors that drive e-

commerce firms’ survival. Current empirical studies on e-commerce (Shareef et al.,

2010, Kaufmann and Wang, 2007) use the same response variable.

I use OLS regressions with a smaller cross-section data set with one date for each firm,

to get a first idea of the relationship of variables used. However, the usage of survival

analysis as the main research method in my work is necessary according to the

literature (Cleves et al., 2010) for the following reasons: observed right-censoring and

left-truncation of the data, and the violation of the normality assumption in the

distribution of time.

Right censoring happens when the event does not take place during the follow-up period

or subjects are lost to follow-up for unknown reasons. Left-truncation or delayed entry

happens when the subject was not observed for a while and then came under

observation. The firms that come to observation even on the first day of study may have

already existed for several months or even years as the data is sourced not from firms

directly, but through a price search engine (the birthdate of firms is the first day a firm

has been listed on geizhals.at). Survival analysis is able to cope with this type of

censored and truncated data (Cleves et al., 2010). I suggest using non-parametric

(Kaplan Meier), semi-parametric (Cox proportional hazards), and competing risk survival

analysis models, which are described further. The following formal description of survival

analysis has been taken from Cleves et al (2010).

Suppose T denotes the length of an e-commerce firm’s survival, and thus is non-

negative. Then the failure function is depicted by:

F(t)=Pr(T<=t)

which is referred to as the cumulative distribution function of survival time. The failure

function then denotes the probability that a random variable T is smaller than some

values t. Reversing the failure function derives the survival function, which depicts the

probability that the event does not occur prior to t.

The survivor function is then given by:

S(t)=1-F(t)=1-Pr(T<=t)=Pr(T>t)

f(t) the density function can be obtained from F(t):

f(t)=dF(t)/dt=d(1-S(t))/dt

Setting the density function f(t) in relation to the survivor function S(t) gives the hazard

function h(t). The hazard function depicts the probability that the event of interest, failure

of e-tailer, occurs conditional upon the fact that it has survived until that point in time:

h(t)=f(t)/S(t)=f(t)/(1-F(t))

Semi-parametric survival analysis assumes that firm i’s hazard rate at time t is related to

a nonparametric baseline hazard rate that is determined by firm age and a parametric

part which is determined by a set of time-varying explanatory variables.

The model can be parameterized as:

h(t|xi) = h0(t)exp(xi βx)

where h0(t) is the baseline hazard for firm i at age t. xi is the vector of time-varying

explanatory variables for firm i at age t, and βx is the vector of parameters to estimate.

Thus, the hazard function depends on time t and explanatory variables. The explanatory

variables in my case are different characteristics of e-commerce firms. Therefore, the

underlying model will help me to identify those statistically significant factors that either

increase or decrease the probability of online firms’ failure.

Further on I also apply Fine and Gray’s competing risks model, as there are different

types of failure present in the data set (status 1, 2, 3). Using Cox proportional hazards

model would produce biased or uninterpretable results by overestimating risks of failure

in the presence of different excluding types of failure (Pintilie, 2011).

Competing risks unlike classical survival analysis is dealing with cumulative incidence

function and cause-specific hazard. Cause-specific hazard is the instantaneous risk of

failure from a specified cause given that no failure from any cause has yet occurred.

Cumulative incidence function (CIF) at time t for cause i is the probability of failing from

cause i before time t (Cleves et al, 2010):

CIFi(t)=P(T<=t and failure from cause i)

The Fine and Gray model is a modification of the Cox proportional hazards model, so

the assumption of proportionality of hazards holds. This model does allow for the

inclusion of excluding failure types and time-varying covariates. The technical

modification consists of keeping the competing events observations in the risk set with a

diminishing weight.

Stata routine was used to compute Fine and Gray competing risks model.

6. ESTIMATION RESULTS

6.1 OLS MODEL: FIRST INSIGHT OF THE DATA

In order to get first comprehension of the data and the relations among variables I have

decided to use Ordinary Least Squares (OLS) method. I have constructed a cross-

section data set by keeping the first observation for every firm in the original data set.

Tabel 1 shows empirical results of an OLS regression. Firms’ lifespan in days,

calculated as difference between firms’ birth date and failure date, is used as dependent

variable. If a firm has not failed during the whole observation period lifespan is

calculated as the difference between the birth date and the last day of observation

period, which is the 9th of September 2012.

According to the results, statistically significant variables (the ones which are not likely to

have occurred purely by chance and thereby allow rejecting the null hypothesis of no

relationship between the measured variables), which explain around 20% of the

variation in the lifespan of e-firms, are: pick-up, the dummy variables for consumer

rating (Rating_good, Rating_average, and Rating_bad; very good rating represents the

base group), Eurolabel, percentage of branded products, frequency of price changes,

percentage of products immediately and shortly available and number of visitors.

All the other independent variables are statically insignificant, meaning that products

HHI, number of products offered, relative price, shipping costs, menu cost time and

average price of clicked products do not influence lifespan of e-commerce firms.

As stated in the OLS results online firms that offer a pick-up possibility exist 261 days

longer than firms that do not. Being multi-channel and having some point of physical

contact with clients has a positive impact on survival of e-commerce.

As expected the Eurolabel increases firms’ lifespan by 510 days, compared to firms that

do not possess that seal of quality. This result suggests importance of external seals of

quality and reputation in e-commerce survival. Increasing branded products offered by

one percentage point decreases e-firms’ lifespan by 121 days, meaning e-tailers are

better off selling unbranded products as e-shoppers seem to prefer not to pay the “brand

fee”. A one percentage point increase in the number of price changes decreases e-

commerce firms’ lifespan by almost 1000 days. This variable is not only statistically

significant at the 1% level, but also its’ practical significance (Wooldridge, 2012) is

impressive. It indicates that firms which change prices less often could live up to almost

3 years more.

Increasing available and shortly available products by one percentage point increases e-

tailers’ lifespan by about 140 days. Firms with ratings good, average and bad all have

longer lifespans than firms with very good ratings. This counter-intuitive result could be

explained by the fact that firms that invest a lot in the quality of their service and receive

very good ratings are unable to cover their higher costs, so their lives are shorter than

those of firms with worse ratings. However it can also be caused by the fact that OLS is

not the proper method to answer my research question.

6.2 KAPLAN-MEIER. GROUP COMPARISONS.

Kaplan Meier is the simplest way of computing survival over time and this method is

often used to find difference in survival probabilities of different groups of subjects under

observation. It is a non-parametric estimate of the survivor function S(t). In the following

section, using Kaplan Meier method, I compare survivor functions of different types of e-

shops to test if there is a difference in the survival pattern depending on such e-shops’

characteristics as sale channels, possession of Eurolabel or number of products offered.

The graphical outcome of Kaplan Meier is a survival curve, which shows survival

probability of subjects under observation for any point in time, which is plotted on the X

axis (Geol et al., 2010).

Figure 3 shows survival curves for single- and multi-channel e-tailers, with time in days

on the X axis and cumulative survival probability on the Y axis. In the lower section of

the graph number of e-tailers at risk and number of failure events (in brackets) by

groups are shown.

0.00 0.25 0.50 0.75 1.00

47 38(0) 34(0) 28(0) 18(0) 0(0)Eurolabel 1049 591(241) 375(90) 231(46) 140(21) 0(14) No Eurolabel

Number at risk

0 500 1000 1500 2000 2500

analysis time

No Eurolabel Eurolabel

Kaplan-Meier survival estimates

0.00 0.25 0.50 0.75 1.00

660 450(143) 317(55) 209(34) 125(17) 0(10)Multi-channel 436 179(98) 92(35) 50(12) 33(4) 0(4)Single-channel

Number at risk

0 500 1000 1500 2000 2500

analysis time

Single-channel Multi-channel

Kaplan-Meier survival estimates

Figure 3. Kaplan-Meier: single- vs. multi-channel e-tailers.

Figure 4. Kaplan-Meier: Eurolabel vs. no Eurolabel e-tailers.

According to Kaplan-Meier survival estimates for single- and multi-channel retailers,

cumulative probability to survive is higher during the whole observation period for multi-

channel retailers with pick-up possibility. This means that additional operational

experience and diversified sales channels provide an advantage for e-commerce firms’

survival.

Kaplan Meier estimates for e-tailers with and without Eurolable show that those firms

that possess seal of quality do not fail during the whole observation period, and firms

that do not have a Eurolabel have lower cumulative probability to survive during the

whole observation period (Figure 4).

Figure 5 shows cumulative survival functions for firms with absolute number of offered

products above the median (“Big”) and below the median (“Small”). During the first year

both groups have almost equal probabilities to survive, after the first year broader

assortment is associated with higher cumulative survival probability. This advantage

increases with time.

Figure 5. Kaplan-Meier: small vs. big e-tailers (by number of products offered).

Kaplan Meier estimates represent non-parametric survival analysis and do not provide

information on the impact of covariates. So further on I present results for competing

risks analysis, which is semi-parametric and also allows for different excluding types of

failures.

6.3 COMPETING RISKS: MAIN RESEARCH METHOD AND RESULTS

Further on I apply competing risks analysis. This method is necessary in the presence of

several excluding types of failures, as traditional survival analysis model like Cox can

produce biased results and overestimate the effect of coefficients.

I use competing risks model which was developed by Fine and Gray – it is a method for

regression analysis that models the hazard that corresponds to the cumulative incidence

function. This model is becoming widely used by researchers and is now available in all

the major software environments (Kuk and Varadhan, 2013).

Full panel dataset providing observations on 1096 online e-tailers for the period of 6

years is used for survival analysis (Kaplan Meier, Competing risks and Cox model). For

this model the firms with status 1 are considered as main failure cause and status 2 and

3 as competing.

Tabel 2 shows empirical results for the estimation of a competing risks model from Fine

and Gray. The results suggest that offering clients a pick-up possibility decreases

hazards by 27%. E-tailers that have physical contact with clients and stay multi-channel

have higher chances to survive than those that do not offer clients pick-up possibility

and stay single-channel.

Austrian e-tailers are better off staying specialized and offering deeper assortment as

increasing concentration in sub-subcategories of products by 1000 points decreases

risks to fail by about 5% (Products HHI). The effect could be also explained by the fact

that specialized e-tailers manage to offer a more targeted choice of products, which

responds better to demand of e-shoppers.

The subhazard rate for Eurolabel is highly significant. The possession of Eurolabel is

associated with an extremely low hazard rate. Online firms that have this external seal of

quality survive throughout the whole observation period. Positive impact of external

seals of quality on survival could be explained by the fact that they increase trust

between clients and e-tailer and decrease uncertainty.

The number of offered products does not determine survival of e-tailers on the Austrian

market, while brand does: increasing branded products offered by 1 percentage point

increases the failure rate on average by 72% (% of branded products). During the

observation period sampled e-tailers are better off selling unbranded products. Table 8

reflects another aspect of brand strategy – brand concentration (Brand HHI). Due to

correlation between brand HHI and the percentage of branded products, it was not

possible to include both in one regression. Results suggest that increasing the HHI of

brand concentration by 1000 is associated with 6% lower failure risks. So less is better

in this case - if Austrain e-tailers chose to sell branded products than higher survival

rates are observed among e-tailers specializing in a certain brand.

Number of price changes per products offered, according to the results, is an important

driver of online retail survival: one percentage point increase in the number of price

changes increases failure risks more than twice. Changing prices too often might be a

sign of e-tailer’s inability to assess markets and chose a consistent price strategy.

Also the time between price changes has an impact on online firms’ survival – one day

increase between price changes is associated with decrease in risk of failure by about

12%. Balanced price change policy seems to be an important survival factor for Austrian

e-commerce market.

Having bad rating compared to having very good rating increases hazards by 34.5%. It

confirms the importance of reputation and feedbacks in internet: providing qualitative

service, keeping customers satisfied and avoiding negative feedbacks is an important

driver of online success.

Also working with the traffic on a current basis is an important driver of e-commerce

survival - increasing number of distinct visitors by 1000, decreases risk of failure by

11.8%.

All the other factors like relative price, shipping costs, and availability of products do not

determine e-commerce survival in Austria.

6.4 ROBUSTNESS TESTS

I have used modifications of the main regression to check if the results are robust. I

have tested whether firms with different price and products offer strategies differ in their

survival pattern. I have also been curious to know whether the world financial crisis of

2007 has influenced the factors of e-commerce success and failure.

Number of products offered

I have divided the sample into two groups with one group of retailers having lower than

the median number of products offered and the second group with the number of offered

products above the median. I suggest there is a correlation between the number of

products offered by an e-tailer and its size, as broader assortment demands higher

costs for stock houses, more personnel, and time resources to maintain and update

information, so I call these two samples the "small firms" and the "big firms”. The first

column of Table 3 provides information on drivers of “small firms” survival: offering pick-

up possibility decreases hazards by 30%; Eurolabel decreases hazards by 100%,

increasing percentage of branded products increases risks by 56.2%, increasing

shipping costs (cash payment) by 1 euro increases risks by about 2%, and increasing

time between price changes decreases failure risks by about 10%. “Small firms” should

pay additional attention to quality of the site and invest in pick-up possibility in order to

increase survival probability.

“Big firms” show a different pattern of survival. The pick-up possibility does not have

impact on their survival, but bad ratings do. Having bad rating compared to very good

increases hazards for “big firms” by around 62%. For “big firms” it is more important to

invest in better ratings than in pick-up option, indicating that reputation counts for them

Eight dummy variables were imputed for missing values with mean values and were checked. One dummy for the number of price

changes turned out to be statistically significant.

more than a personal contact with clients, which is an important driver for “small firms”.

Also the disadvantage of selling a bigger share of branded products is higher for “big

firms”. Number of distinct visitors has an impact on “big firms” survival only, increasing

number of distinct visitors by 1000, decreases hazards by 12.5%.

Price

I have also checked if the main results are still robust for high and low price e-firms. I

have divided the dataset in 2 groups below and above the median average price of

325.1 euro.

According to Table 4, the numbers of price changes and the relative price have only

impact on low price firms’ survival. Low price firms are better off if they stay specialized

in sub-subcategories of products – increasing the products HHI by 1000 decreases

hazards by around 10%. The number of distinct visitors decreases hazards for both high

and low price firms. The advantage of a higher number of visitors is larger for high price

firms. Firms whose average price is above the median can decrease failure hazards by

15% by increasing number of unique visitors by 1000. This can be explained by the fact

that for high price firms bigger loyal customers’ base is more important than for low price

firms. Shipping costs have a negative impact on survival of high price firms, which might

be explained by the fact that shipping costs are significantly higher and might determine

clients’ decisions to commit an order. Also decreasing time between price changes

increases hazards of high price firms.

Age

Table 5 gives an overview on survival rates of old and young firms. For this robustness

check the firms have been divided by birthdate (first appearance on geizhals.at) in 3

categories. The first category includes only all the firms born on the first day of the

observation period, the second group (old firms) consists of firms born between the

second day of the observation period and January 1, 2007 (median birthdate), the third

group (young firms) consists of the firms born after January 1, 2007. Both the second

and the third groups represent inflow samples.

The pick-up option only decreases hazards for young firms. These firms profit from a

small decrease in hazards if they are more specialized. A raise in the percentage of

branded products increases risks for young firms only. The frequency of price changes

effects only the survival of young firms – more frequent price changes increase hazards

by almost 3.5 times. Meaning firms born after January 1, 2007 should be cautious in

setting reasonable prices and change them as less as possible in order to increase

survival probability.

For old firms one additional day between price changes decreases hazards by 16%.

Bad ratings decrease chances on success just for old firms. Probably in earlier days of

e-commerce, when fewer clients have been buying online and the whole online market

has been very new, reputation and trust have been very important drivers, triggering of

order decision and determining survival.

Increase of available products by one percentage point is associated with 53% decrease

in hazards for firms born on or before the first day of observation, April 23, 2006.

Increasing shipping costs payment in advance decreased hazards by 13% for old firms.

But increasing shipping costs payment by cash on delivery increases risks by 9% for old

firms. Average price of clicked products decreases hazards slightly only for firms born on

or before April 23, 2006.

Crisis

The observation period includes the Global Financial crisis of 2007- 2008, so by dividing

data by timestamp in 3 periods before crisis (from April 23, 2006 to August 8, 2007),

during crisis (August 9, 2007 to January 1, 2009) and after crisis (from January 2, 2009

to September 9, 2012), I get an overview on how impacts of factors that determine e-

commerce survival in Austria changed at different time periods (including also the period

of the global financial crisis).

Results are presented in Table 6. The number of unique visitors, Eurolabel and the time

between price changes are the only covariates that show the same impact on e-

commerce survival before, during and after crisis.

Before crisis firms increasing the percentage of branded products by one percentage

point more than doubled their risk of failure. Also before crisis increasing the average

price of clicked products by 1 euro decreases hazards by 1%.

During the crisis increasing the number of offered products by 100 leads to 0.6% higher

risk of failure. This is quite intuitive: the higher are the potential expenses on stocks and

information uploading, the lower are the chances to survive during recession period. The

percentage of branded products does not determine survival of e-commerce in times of

crisis. The pick-up possibility has become an advantage during crisis, also after crisis

firms offering pick-up option are better off as they have 30% lower hazards. Price

change and bad rating have become significant drivers of e-tailers failure only after

crisis. Also increasing specialization on sub-subcategories has become an advantage

after crisis.

Two last columns of Table 6 capture the same idea of pre- and post-crisis times, but

with different time division. For this analysis all the firms have been divided by

timestamp on two groups: the first group represents first half of the observation period

from 2006 to 2009 and the second group – second half of the observation period from

2009 to 2012.

During the first half of the study number of offered products and percentage of branded

products offered show negative impact on e-commerce firms’ survival. Also during the

first half of observation period increasing the percentage of immediately available

products by one percentage point decreases failure risks by 32.6%: during this period

clients might have been sensitive to receiving the products as soon as possible.

During the second half of the observation period increasing products HHI by 1000

decreases failure risks by 7.3%, which means specialized shops have better chances to

survive. The frequency of price changes in the second half of the study has a significant

impact on survival decreasing survival probability by about 5.5 times. The bad rating has

a negative impact on survival increasing failure hazards by 55.9%.

During the whole observation period the pick-up option, Eurolabel and the number of

distinct visitors have positive impact on survival, decreasing failure hazards.

Cox

Cox model is the cornerstone of modern survival analysis; it is widely used and

comfortable to interpret (Lawless, 1998). Different types of failure, like the

disappearance from geizhals.at, are an interesting phenomenon, which cannot be

ignored and should be analyzed. Several modifications of Cox model have been used to

track drivers of different types of failure. However, according to the theory Cox model

results in presence of competing failures can be biased (Cleves et al., 2010). Still

irrespective of the modification of model used, Cox results are similar to main competing

risks model shown in the first column of Table 7.

In the second column only firms with status equal to 1 (no online presence – “full failure”

of e-tailer) are treated as subjects that have experienced an event (failure) during the

observation period, firms with status 2 and 3 are treated as censored. This model

behaves identically to main competing risk model.

The third column shows results for firms with status 1 (“full failure”), 2 (“question marks”),

and 3 (“geizhals death”) are all treated as failure. According to this model having

average rating increases hazards rates by about 37% and increasing percentage of

immediately available products by one percentage point decreases risks by 24%. The

percentage of branded products and shipping costs do not determine e-commerce firms’

survival.

In the fourth column, status 1 and 3 are considered as failure and 2 is censored, in this

modification we check for firms that still have some evidence of online presence, but

have stopped being listed on geizhals.at. This group shows slightly different pattern of

survival. For this type of failure percentage of branded products offered is not

determining survival, bad ratings increase hazards by 56.1%, and additional 1000

unique visitors decrease risks by about 14%. Also increasing the percentage of

immediately available products by 1 percentage point decreases risks by roughly 25%.

The last column displays results, where status 1 and 2 are equal to failure and 3 is

censored. This group shows quite similar results to the competing risk model, which

could mean that “question mark” firms are closer to complete failure than to operating e-

firms.

7. CONCLUSIONS AND LIMITATIONS

According to the data on the full set of 1096 Austrian e-tailers observed during the

period from 2006 to 2012 on the Geizhals.at website important drivers of e-commerce

survival are: pick-up possibility, seals of quality, percentage and concentration of

branded products, number of visitors and price change strategy. Offering pick-up option,

having seals of quality, concentration on one or a few brands and increasing number of

unique visitors have a positive impact on e-commerce survival, while offering branded

products and increasing frequency and size of price changes has a negative impact of

survival of e-commerce. These main results hold regardless of the model and method of

research.

Competing risk survival analysis is chosen as the main research method although

irrespective of the modification Cox model shows similar results. The analysis of

different types of firms and different time periods revealed the following more detailed

results:

 Small firms profit from offering a pick-up possibility; which is in line with the idea

that small business profitability is often built on better customer comprehension

and relations, more contact and intimacy with customers.

 Big firms do not profit from offering a pick-up option, but they do profit from better

ratings as bad ratings increase significantly the risk of failure for them; this result

could be explained by the fact that even if big firms have a point of pick-up they

are, due to higher clients traffic, unable to provide the same kind of personal

approach and profit from it like small firms do, so big firms increase survival

chances by investing in a better image and reputation.

 Low price firms increase hazards by offering branded products and increasing the

frequency of price changes. This indicates quite intuitive idea that clients buying

from low price firms are probably looking for the cheapest option and do not care

about the brand.

 Low price firms can decrease failure hazards by increasing concentration on sub-

subcategories of products, increasing relative price and attracting a higher

number of distinct visitors.

 For old firms getting good ratings is associated with higher survival rates, unlike

for young ones, which could be explained by the fact that in the beginning of the

e-commerce building a reputation was a more important factor defining survival.

 The size of price changes does not affect survival of old firms, but the frequency

of price changes does.

 Bad ratings and higher shipping costs have negative impact on survival of old

firms, but no impact on young firms.

 Increasing number of unique visitors decreases hazards for old firms, but not for

young ones.

 Firms operating in crisis face higher risk with wider assortment, they also increase

hazards by staying single-channel. This means that in crisis times expenses on

broad assortment should be cut and better comprehension, contact with

customers is needed.

Further research could have included more financial information or information on firms’

the corporate structure (e. g. information about the CEO, number of employees).

Also checking the e-commerce shops’ availability of mobile versions or mobile

application could be interesting. Today, mobile commerce takes up an increasingly

important role in e-commerce. According to the study that was conducted by the

Handelsverband, mobile commerce is booming in Austria, as there has been 25% more

spendings via smart phones in 2013 than in previous year (Ecommerce news, 2014). So

offering a mobile version of an online shop can be a significant advantage in terms of

higher clients’ traffic. E-shops also offer apps (applications) that can be downloaded for

free in the appstores of different operation system provider (e. g. Apple, Google). Apps

have a faster and easier access to online shops and aim to create a base of loyal

customers. So I suggest that having an application or mobile version might significantly

increase the chances of e-tailers to survive.

Geizhals.at does not provide any information on e-tailers that sell everyday clothes or

accessories, which is an important part of the e-commerce market. For example, in 2012

in the US the sales’ share of apparel and accessories in total e-retail sales was 18.3%.

And it is expected to grow to 20.2% by 2016 (Statista, 2015). Testing e-tailers that

specialize in apparel and accessories could contribute to future research.

Comparing results of this work with other European markets as well as with e-tailers

from US, CIS or Asian countries could also be an interesting path for further research.

Many international retailers like Zara or Apple have online shops in countries all over the

world and finding out whether different survival strategies corresponding to different

markets exist could provide value added not only for researchers, but also increase

efficiency of country-specific strategies of e-tailers.

REFERENCES

Agarwal R., Gort M., 2002. “Product Life Cycles and Firm Survival.” American Economic

Review, 92(2): 184-190.

Audretsch D., Mahmood T., 1995. “New Firm Survival: New Results Using a Hazard

Function.” The Review of Economics and Statistics, 77(1): 97-103.

Banerjee S., Kauffman R.J., Wang B., 2007. “Modeling Internet firm survival using

Bayesian dynamic models with time-varying coefficients.” Electronic Commerce

Research and Applications Journal, 6(3): 332-342.

Cleves M., Gould W., Gutierrez R.G., Marchenko Y., 2010. “An Introduction to Survival

Analysis Using Stata.” Stata Press, College Station, Texas 77845, p. 1-391.

Cosgun V., Dogerlioglu O., 2012. “Critical Success Factors Affecting e-commerce

Activities of Small and Medium Enterprises.” Information Technology Journal, 11(12):

1664-1676.

Emarketer.com, 2014. “Global B2C Ecommerce Sales to Hit $1.5 Trillion This Year

Driven by Growth in Emerging Markets.” http://www.emarketer.com/Article/Global-B2C-

Ecommerce-Sales-Hit-15-Trillion-This-Year-Driven-by-Growth-Emerging-

Markets/1010575. [Online; accessed January 02, 2015].

Entrepreneur.com, 2014 “How to raise and lower you prices.”

http://www.entrepreneur.com/article/66010. [Online; accessed January 02, 2015].

Hackl F., Kügler A., Winter-Ebmer R., 2011. „Reputation and Certification in Online

Shops.“ Working Paper Series of the Department of Economics at the University of Linz,

Austria.

Hannan M.T., Freeman J., 1977. “The population ecology of Organizations.” American

Journal of Sociology, 82(5): 929-964.

Kauffman R.J., Wang B., 2002. “Duration in the digital economy: empirical bases for the

survival of internet firms.” Proceedings of the 36th Annual Hawaii International

Conference.

Kuk D., Varadhan R., 2013. “Model selection in competing risks regression.” Statistics in

Medicine, 32(18): 3077-3088.

Lawless J., 1998. “Parametric models in survival analysis.” Encyclopedia of Biostatistics,

3254-64.

Levins R., 1962. “Theory of fitness in a heterogeneous environment.” American

Naturalist, 96(4): 361-378.

Loiacono E.T., Watson R.T., Goodhue D.L., 2002. “Webqual: A Measure of Website

Quality,” Marketing Theory and Applications, American Marketing Association, Chicago,

p. 13.

Mahajan V., 2002. “The Dot.com Retail Failures of 2000: Were There Any Winners?”

Journal of the Academy of Marketing Science, 30(4): 474-486.

Neslin S. A., Shankarb V., 2009. ”Key Issues in Multichannel Customer Management:

Current Knowledge and Future Directions.” Journal of Interactive Marketing, 23(1): 70–

81.

Nikolaeva R., 2007. “The Dynamic Nature of Survival Determinants in E-commerce.”

Journal of Academy of Marketing Science, 35(4): 560-571.

Osterwalder A., “Business survival toolkit.” http://business-survival-toolkit.co.uk/stage-

one/business-model-analysis/business-model-generation-canvas. [Online; accessed

November 15, 2014].

Parasuraman A., Zeithaml V.A., 2005. “E-S-QUAL a Multiple-Item Scale for Assessing

Electronic Service Quality.” Journal of Service Research, 7(3): 213-233.

Pintilie M., 2011. “An Introduction to Competing Risks Analysis.” Revista Espanola de

Cardiologia, 64(7): 599-605.

Randall T., Ulrich K., 2001. “Product variety, supply chain structure, and firm

performance: analysis of the US bicycle industry.” Management Science, 47(12): 588-

604.

Rembor E., 2010. ”307 ways to build a successful Web-shop.” Arpe Publication, London

UK, p.12-48.

Shareef Mahmud A., 2008. ” Studying the Application of Survival Analysis: Modeling the

Survivability of E-Commerce.” International Journal of Business Insights and

Transformation, 8(1): 1-34.

Statista.com, 2015. “Share of apparel and accessories sales in total U.S. e-retail sales

from 2010 to 2016.” http://www.statista.com/statistics/281594/share-of-apparel-and-

accessories-sales-in-total-us-e-retail-sales/. [Online; accessed January 02, 2015].

Van Paesschen H., Vis D., 2014. “Mcommerce is booming in Austria.“

http://ecommercenews.eu/mcommerce-is-booming-in-austria/. [Online; accessed

November 18, 2014].

Varadarajan P., Yadav M., 2002. “Marketing Strategy and the Internet: An Organizing

Framework.” Journal of the Academy of Marketing Science, 30(4): 296-312.

Wolfinbarger M., 2003. “Etailq: Dimensionalizing, Measuring and Predicting Etail

Quality.” Journal of Retailing, 79(3): 183-198.

Wooldridge J., 2012. “Introductory econometrics: A modern approach.” Cengage

Learning.

APPENDIX

Table 1. Determinants of e-commerce firms’ lifespan in Austria.

Dependent variable: e-

tailers’ lifespan (days)

Pick_up

261.56

(46.36)***

Products HHI

-1.39

(8.85121)

Rating_good

206.05

(74.68)***

Rating_avg

277.06

(96.09)***

Rating_bad

210.93

(75.11)***

Eurolabel

510.62

(107.33)***

Number of products offered

-0.18

(0.44)

% of branded products

-121.53

(70.57)*

Relative price

365.68

(224.41)

Shipping

6.28

(4.01)

Shipping_cash_payment

0.25

(4.28)

Number of price changes

-967.53

(213.72)***

Menu_cost_time

17.16

(15.22)

Menu_cost_time2

-0.86

(0.49)*

Menu_cost_height

-34.28

(83.00)

Immediately available

146.41

(60.57)**

Shortly available

141.83

(72.39)*

Number of visitors

21.81

(5.39)***

Click_price_avg

0.00

(0.03)

Dum_shipping_vk

5.83

* p<0.1; ** p<0.05; *** p<0.01

(48.51)

Dum_rating

-424.89

(90.14)***

Dum_price_change

6,108.29

(15,211.54)

Dum_click_price_avg

-9.99

(272.29)

_cons

319.09

(261.07)

0.20

1,096

Table 2. Drivers of e-commerce success and failure.

Robust z-statistics in parentheses

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

Drivers of e-commerce success and failure in

Austria

(2006-2012)

Pick_up

0.736***

(-2.792)

Products HHI

0.953**

(-2.003)

Eurolabel

0.000***

(-112.596)

Number of products offered

1.001

(1.180)

% of branded products

1.719***

(3.215)

Relative price

0.702

(-0.642)

Number of price changes

2.279*

(1.873)

Menu_cost_height

0.879

(-0.736)

Menu_cost_time

0.882***

(-3.760)

Menu_cost_time2

1.004***

(3.748)

Rating_good

1.049

(0.270)

Rating_avg

1.154

(0.637)

Rating_bad

1.345*

(1.772)

Immediately_available

0.894

(-0.745)

Shorty_available

0.837

(-1.000)

Shipping

0.997

(-0.347)

Shipping_cash

1.018*

(1.933)

0.882***

(-3.604)

Click_price_avg

0.999

(-1.148)

Dum_price_change

0.069***

(-5.110)

Observations

29,401

Table 3. Drivers of e-commerce success and failure. Small vs. big firms.

Robust z-statistics

in parentheses

*** p<0.01, **

p<0.05, * p<0.1

Drivers of e-commerce success and failure

Small firms

Big firms

Pick_up

0.700***

0.764

(-2.645)

(-1.433)

Products HHI

0.961

0.826*

(-1.510)

(-1.814)

Rating_good

1.086

1.042

(0.367)

(0.140)

Rating_avg

1.154

1.138

(0.448)

(0.383)

Rating_bad

1.143

1.616*

(0.568)

(1.773)

Eurolabel

0.000***

(-105.811)

(-65.388)

Number of products offered

0.976

1.001

(-0.249)

(0.453)

% of branded products

1.562**

2.205**

(2.309)

(2.251)

Relative price

1.004

0.097*

(0.201)

(-1.816)

Shipping

1.003

0.985

(0.269)

(-0.932)

Shipping_cash_payment

1.021*

1.002

(1.949)

(0.113)

Number of price changes

2.301

0.581

(1.608)

(-0.450)

Menu_cost_time

0.901**

0.868**

(-2.394)

(-2.163)

Menu_cost_time2

1.003***

1.004*

(2.599)

(1.676)

Menu_cost_height

0.609*

0.995

(-1.826)

(-0.497)

Immediately available

0.907

0.629

(-0.535)

(-1.603)

Shortly available

0.662

0.998

(-1.633)

(-0.008)

Number of visitors

0.846

0.885***

(-1.057)

(-3.442)

Click_price_avg

1.000

(-0.740)

(-0.684)

Dum_shipping_vk

1.082

0.775

(0.529)

(-1.232)

Dum_price_change

0.070***

7.764

(-3.872)

(1.053)

Dum_menu_height

8.276e+40*

(1.890)

Observations

14,709

14,692

Table 4. Drivers of e-commerce success and failure. Low price vs. high price firms.

Drivers of e-commerce

success and failure

Low price

(<=325.1€)

High price

(>325.1€)

Pick_up

0.744**

0.742*

(-2.056)

(-1.696)

Products HHI

0.909***

1.016

(-2.878)

(0.456)

Rating_good

1.090

1.059

(0.354)

(0.221)

Rating_avg

1.007

1.352

(0.020)

(0.926)

Rating_bad

1.308

1.331

(1.170)

(1.143)

Eurolabel

0.000***

(-92.325)

(-69.931)

Number of products offered

0.999

1.002

(-0.245)

(1.556)

% of branded products

1.772**

1.837**

(2.446)

(2.388)

Relative price

0.334*

1.010

(-1.666)

(0.782)

Shipping

0.996

1.004

(-0.137)

(0.414)

Shipping_cash_payment

1.024

1.020*

(1.141)

(1.870)

Number of price changes

2.729*

1.495

(1.776)

(0.558)

Menu_cost_time

0.933

0.818***

(-1.558)

(-4.050)

Menu_cost_time2

1.003*

1.006***

(1.895)

(3.604)

Menu_cost_height

0.710

0.957

(-0.928)

(-0.266)

Immediately available

0.879

0.692

(-0.647)

(-1.477)

Shortly available

0.737

0.836

(-1.205)

(-0.689)

Number of visitors

0.911**

0.855***

(-2.184)

(-3.116)

Click_price_avg

1.000

(0.556)

(-1.034)

Dum_shipping_vk

0.833

1.123

(-0.996)

(0.648)

Dum_rating

0.897

0.802

(-0.352)

(-0.687)

Robust z-statistics in parentheses

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

Dum_price_change

0.085***

7,248.556

(-4.168)

(0.294)

Dum_click_price_avg

0.000***

1.330

(-27.132)

(0.324)

Observations

14,700

14,701

Table 5. Drivers of e-commerce success and failure. Old vs. young firms (by birthdate).

Inflow sample

Drivers of e-commerce success and

failure

Born on

23.04.2006

Born

24.04.2006-

07.01.2007

Born after

07.01.2007

Pick_up

0.728

1.396

0.682***

(-1.243)

(0.816)

(-2.988)

Products HHI

0.976

0.922

0.946**

(-0.398)

(-0.614)

(-2.025)

Rating_good

0.616

1.629

1.177

(-1.191)

(0.864)

(0.762)

Rating_avg

1.478

0.649

0.974

(1.006)

(-0.484)

(-0.077)

Rating_bad

1.084

2.807*

1.269

(0.223)

(1.956)

(1.128)

Eurolabel

0.000***

(-43.124)

(-21.559)

(-74.671)

Number of products offered

0.996*

1.004

1.002

(-1.660)

(1.087)

(1.579)

% of branded products

2.641**

1.005

1.785***

(2.325)

(0.007)

(2.900)

Relative price

0.499

1.800

0.844

(-0.556)

(0.305)

(-0.284)

Shipping

0.975

0.875**

1.006

(-1.221)

(-2.039)

(0.542)

Shipping_cash_payment

1.015

1.090**

1.014

(0.680)

(2.216)

(1.301)

Number of price changes

0.684

1.146

3.450**

(-0.333)

(0.116)

(2.509)

Menu_cost_time

0.868*

0.835**

0.900**

(-1.719)

(-2.097)

(-2.575)

Menu_cost_time2

1.003

1.005*

1.004***

(1.199)

(1.750)

(2.912)

Menu_cost_height

0.995

0.942

0.879

(-0.050)

(-0.382)

(-0.544)

Immediately available

0.490*

0.817

1.038

(-1.878)

(-0.404)

(0.212)

Shortly available

0.739

0.474

0.939

(-0.772)

(-1.333)

(-0.289)

Number of visitors

0.890**

0.931

0.905

(-2.181)

(-1.198)

(-1.542)

Click_price_avg

0.999*

0.999

1.000

(-1.667)

(-1.111)

(-0.705)

Dum_shipping_vk

0.740

1.513

1.032

(-1.094)

(0.897)

(0.218)

Dum_rating

0.679

2.154

0.918

(-0.921)

(0.922)

(-0.259)

Dum_price_change

6.879

123,194.57

0.110***

(0.095)

(0.406)

(-4.282)

Dum_click_price_avg

0.000***

1.416

(-22.448)

(-9.445)

(0.447)

Observations

11,286

3,434

14,681

Robust z-statistics in parentheses

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

Table 6. Drivers of e-commerce success and failure. Crisis (by timestamp).

Drivers of e-commerce

success and failure

Before

crisis

Crisis

After crisis

2006-2009

2009-2012

Pick_up

1.097

0.512***

0.708**

0.743**

0.697**

(0.325)

(-2.734)

(-2.506)

(-1.969)

(-2.346)

Products HHI

0.987

0.989

0.937**

0.979

0.927**

(-0.194)

(-0.189)

(-2.277)

(-0.575)

(-2.442)

Rating_good

1.332

1.002

1.005

0.899

1.237

(0.598)

(0.006)

(0.024)

(-0.421)

(0.842)

Rating_avg

2.275

1.001

0.980

0.940

1.660

(1.560)

(0.003)

(-0.065)

(-0.200)

(1.546)

Rating_bad

1.652

0.980

1.500**

1.278

1.559*

(1.014)

(-0.047)

(1.974)

(1.060)

(1.830)

Eurolabel

0.000***

(-44.019)

(-45.026)

(-84.535)

(-65.185)

(-91.424)

Number of products offered

1.000

1.006***

1.000

1.004***

0.998

(0.025)

(2.686)

(-0.053)

(2.592)

(-0.865)

% of branded products

2.244*

1.471

1.699**

2.169***

1.379

(1.921)

(1.079)

(2.506)

(3.277)

(1.342)

Relative price

0.152

0.480

0.918

0.480

1.002

(-1.025)

(-0.638)

(-0.150)

(-0.923)

(0.102)

Shipping

0.995

1.017

1.007

1.003

1.013

(-0.184)

(1.279)

(0.750)

(0.253)

(1.179)

Number of price changes

0.940

0.703

3.414**

0.963

5.540***

(-0.069)

(-0.348)

(2.247)

(-0.057)

(3.139)

Menu_cost_ time

0.771***

0.883*

0.910**

0.876***

0.875***

(-3.465)

(-1.681)

(-2.207)

(-2.900)

(-2.734)

Menu_cost_time2

1.007***

1.004

1.003***

1.004***

(2.596)

(1.556)

(2.579)

(2.716)

(2.895)

Immediately available

0.527

0.765

0.986

0.674*

1.093

(-1.621)

(-0.720)

(-0.077)

(-1.787)

(0.444)

Shortly available

0.568

0.795

0.908

0.708

0.955

(-1.289)

(-0.616)

(-0.428)

(-1.403)

(-0.184)

Number of visitors

0.794**

0.885*

0.908**

0.877***

0.852**

(-2.136)

(-1.939)

(-2.052)

(-3.267)

(-2.111)

Click_price_avg

0.999*

1.000

(-1.653)

(-0.066)

(-0.918)

(-0.912)

(-0.626)

Dum_shipping_vk

1.042

0.680

1.008

0.857

0.994

(0.142)

(-1.001)

(0.058)

(-0.775)

(-0.037)

Dum_rating

0.766

1.466

0.799

1.252

0.489**

(-0.572)

(0.818)

(-0.742)

(0.752)

(-2.272)

Dum_price_change

2.897***

1.855*

2.182***

2.361***

2.037***

(3.005)

(1.841)

(4.677)

(4.293)

(3.788)

Dum_click_price_avg

0.000***

2.346

0.000***

3.509*

(-13.818)

(-17.439)

(1.063)

(-26.133)

(1.658)

Observations

4,284

5,670

19,447

14,824

14,577

Robust z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 7. Drivers of e-commerce success and failure. Competing risks vs. Cox proportional hazards.

Drivers of e-commerce

success and failure

Competing

risk

Cox

(1)

Cox

(1,2,3)

Cox

(1,3)

Cox

(1,2)

Pick_up

0.736***

0.676***

0.692***

0.64***

0.734***

(-2.792)

(-3.614)

(-4.292)

(-4.78)

(-3.123)

Products HHI

0.953**

0.952**

0.971*

0.975

0.951**

(-2.003)

(-2.200)

(-1.764)

(-1.390)

(-2.489)

Eurolabel

0.000***

0.000

0.332***

0.350***

0.037***

(-112.596)

(-0.000)

(-3.732)

(-3.397)

(-3.289)

Number of products offered

1.001

0.999

1.000

0.999

(1.180)

(0.608)

(-0.097)

(0.530)

(-0.070)

% of branded products

1.719***

1.571***

1.175

1.204

1.444**

(3.215)

(2.788)

(1.270)

(1.346)

(2.546)

Relative price

0.702

0.688

0.904

0.875

0.751

(-0.642)

(-0.710)

(-0.284)

(-0.342)

(-0.613)

Number of price changes

2.279*

2.894**

3.322***

2.249*

4.478***

(1.873)

(2.248)

(3.353)

(1.954)

(3.810)

Menu_cost_height

0.879

0.874

0.881

0.875

0.879

(-0.736)

(-0.717)

(-0.896)

(-0.821)

(-0.819)

Menu_cost_time

0.882***

0.869***

0.881***

0.873***

0.881***

(-3.760)

(-4.073)

(-4.567)

(-4.627)

(-3.998)

Menu_cost_time2

1.004***

(3.748)

(3.999)

(4.364)

(4.392)

(3.943)

Rating_good

1.049

1.078

1.083

1.119

1.045

(0.270)

(0.421)

(0.518)

(0.690)

(0.261)

Rating_avg

1.154

1.186

1.372*

1.293

1.297

(0.636)

(0.781)

(1.757)

(1.326)

(1.307)

Rating_bad

1.345*

1.470**

1.614***

1.561***

1.547***

(1.772)

(2.305)

(3.409)

(2.981)

(2.825)

Immediately_available

0.894

0.804

0.760**

0.747**

0.816

(-0.745)

(-1.514)

(-2.414)

(-2.396)

(-1.557)

Shorty_available

0.837

0.857

0.930

0.861

0.950

(-1.000)

(-0.919)

(-0.552)

(-1.053)

(-0.344)

Shipping

0.997

0.992

1.001

(-0.347)

(-0.309)

(-0.371)

(-0.834)

(0.136)

Shipping_payment_cash

1.018*

1.016*

1.010

1.016*

1.010

(1.933)

(1.720)

(1.308)

(1.894)

(1.106)

Number of visitors

0.882***

0.878***

0.867***

0.890***

(-3.604)

(-3.605)

(-4.313)

(-4.324)

(-3.594)

Click_price_avg

0.999

1.000

0.999

1.000

(-1.148)

(-0.865)

(0.040)

(-0.717)

(0.051)

Dum_price_change*

0.069***

0.094**

0.137*

0.120**

0.108**

(-5.110)

(-2.219)

(-1.923)

(-2.039)

(-2.109)

Observations

29,401

Robust z-statistics in parentheses

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

Table 8. Drivers of e-commerce success and failure. Brand concentration.

Robust z-statistics in parentheses

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

Drivers of e-commerce success and failure

Pick_up

0.724***

(-2.935)

Products HHI

0.970

(-1.233)

Rating_good

1.039

(0.213)

Rating_avg

1.129

(0.538)

Rating_bad

1.324*

(1.673)

Eurolabel

0.000***

(-110.056)

Number of products offered

1.001

(0.745)

% of branded products

1.258

(0.986)

Brand HHI

0.946**

(-2.124)

Relative price

0.686

(-0.683)

Shipping

0.996

(-0.383)

Shipping_cash_payment

1.019**

(1.964)

Number of price changes

2.339*

(1.911)

Menu_cost_time

0.882***

(-3.756)

Menu_cost_time2

1.004***

(3.838)

Menu_cost_height

0.892

(-0.693)

Immediately available

0.875

(-0.879)

Shortly available

0.833

(-1.025)

Number of visitors

0.876***

(-3.707)

Click_price_avg

1.000

(-0.968)

Dum_shipping_vk

0.954

(-0.388)

Dum_price_change

0.079***

(-4.786)

Observations

29,401

DESCRIPTIVE STATISTICS

Survival analysis

mean

min

max

Yearofaward

2007.426

2.909553

2001

2014

Alive_2013

.5724295

.4947346

Firm_lifespan_days

1596.582

722.4596

2445

Data_id

14959.49

8645.158

30123

Offer

2640.606

7244.904

145890

Products_hhi

2040.899

2491.651

10000

Brand_hhi

4951.882

3264.819

554

10000

Percentage_of_branded_products

.4694762

.3196375

Price_avg

541.9763

10217.65

1000000

Price_rel

1.008478

3.057333

.265758

524.859

Immediately_availible

.3884181

.3780242

Shortly_availible

.2170051

.3270651

Shipping_at

7.497715

6.377629

94.28

Shipping_cash

ноя.67

5.828549

63.25

Click_cnt

3367.602

14928.57

529520

Click_price_avg

486.4356

1006.754

77944.43

Clicks_hhi

2798.755

2660.877

105

10000

1870.551

6074.622

154852

Price_changes

.0564149

.0859213

.0001

Menu_cost_sync

.0517711

.090804

Menu_cost_height

185.387

14583.04

-1

19574

Menu_cost_time

11.75384

5.143988

Pick_up_se~e

.7326962

.4425598

Firms

541.7825

322.6867

1096

Timest

18198.89

643.4374

16930

19245

Reap

.2384273

.5487966

Eurolabel

.0760858

.2651399

Birth

17421.72

592.9469

16914

19238

Death_date

18575.7

561.5331

16939

19265

End_period

18198.5

643.5296

16930

19245

Last_update

19018.43

530.4751

16939

19360

Dum_shipping_at

.262202

.4398394

Dum_shipping_cash

.3010102

.4587049

Dum_rating

.275875

.4469617

Dum_click_avg

.0023469

.0483883

Dum_cookies

.0023469

.0483883

Dum_price_changes

.166593

.3726184

Dum_price_rel

.000102

.010101

Dum_menu_height

.166695

.3727097

Dum_menu_time

.166593

.3726184

Rating1

.1999592

.3999762

Rating2

.2009455

.4007142

Rating3

.3820618

.4858998

Rating4

.2170334

.4122326

Menu_time2

164.6125

149.4573

961

Offer_100

26.40606

72.44904

.01

1458.9

Brand_hhh_1000

4.951882

3.264819

.554

Product_hhi_1000

2.040899

2.491651

.056

Cookies_1000

1.870551

6.074622

.001

154.852

Status

.7250434

1.015882

Last

18947.27

500.1665

16930

19245

Event

.0303391

.2138836

_St

.0140131

.1175467

_Origin

17421.72

592.9469

16914

19238

777.1719

600.9584

2331

_T0

744.0194

600.6066

2300

OLS

mean

min

Max

Yearofaward

2008.319

3.356411

2001

2014

Alive_2013

.4051095

.4911373

Firm_lifespan_days

946.1651

776.7312

2445

Data_id

10339.28

9361.019

30121

Offer

1489.195

4211.949

37084

Products_hhi

2980.914

3139.605

10000

Brand_hhi

5527.364

3467.545

673

10000

Percentage_of_branded_products

.4614153

.3512858

Price_avg

416.7624

505.8167

8064.05

Price_rel

.9852677

.3077142

.396448

1617934

Immediately_availible

.3670381

.4049377

Shortly_availible

.1767819

.3233706

Shipping_at

7.15026

5.74492

74.2

Shipping_cash

10.76721

6.173862

63.25

Click_cnt

895.6113

3497.759

82328

Click_price_avg

467.8307

857.6803

16031.37

Clicks_hhi

3682.672

3036.058

146

10000

574.2322

1444.08

21181

Price_changes

.0897949

.1372469

.0002889

Menu_cost_sync

.0830627

.1405349

Menu_cost_height

40.18992

75.82706

-1

18.320

Menu_cost_time

15.80186

6.626445

Pick_up_se~e

.6021898

.4896693

Firms

557.3923

322.4126

1114

Timest

17814.17

739.8495

16930

19245

Reap

.3001825

.6105914

Eurolabel

.0428832

.2026863

Birth

17761.79

734.0367

16914

19238

Death_date

18272.58

675.8721

16939

19265

End_period

17811.5

741.4174

16930

19245

Last_update

18708.07

741.9403

16939

19360

Dum_shipping_at

.3731752

.4838689

Dum_shipping_cash

.3448905

.4755496

Dum_rating

.6213504

.485272

Dum_click_avg

.0164234

.127155

Dum_cookies

.0164234

.127155

Dum_price_changes

.2180657

.4131208

Dum_price_rel

.0009124

.0302061

Dum_menu_height

.2189781

.4137425

Dum_menu_time

.2180657

.4131208

Rating1

.1313869

.3379772

Rating2

.1140511

.3180184

Rating3

.665146

.4721549

Rating4

.0894161

.2854736

Menu_time2

293.5685

236.1429

961

Offer_100

14.89195

42.11949

.01

370.84

Brand_hhh_1000

5.527364

3.467545

.673

Product_hhi_1000

2.980914

3.139605

.073

Cookies_1000

.5742322

1.44408

.001

21.181

0 views·58 pages

E-commerce: Drivers of Success and Failure PDF Free Download

E-commerce: Drivers of Success and Failure PDF free Download. Think more deeply and widely.

Uploaded by im_daniel on 4/9/2026

/58

100%