Using Software Stacks to Explain
Complementarities: The Case of Mergers
and Acquisitions in the Software Industry
Lucia Silva and Bala Iyer
Boston University
[email protected], [email protected]
HICSS-39, January 2006
1
The purpose of the paper is:
1)
To test the hypothesis that there is value in equity
participation between companies that produce different
components of a complementary network;
Definition of complementarity by Milgrom and Roberts
(1995): “the whole is more than the sum of the parts”, i.e.,
the returns obtained from combining the activities are
greater than the sum of the returns of both activities in
isolation.
2)
To empirically validate the software stack.
2
The Stack
Service
Application software
Middleware Services
Systems Software
Hardware
3
Industry Stack
Services
(IT consulting, systems integration, outsourcing, training & education,
maintenance)
Application software
(CRM package, SCM package, BI, HR, engineering & design, e-biz)
Middleware Services
(systems management, application & transaction servers, collaboration & messaging
database)
Systems
(OS, file sharing, terminal access)
Hardware
4
Stack

The industrial organization of the software industry can be
structured according to an approach imported from the software
architecture, commonly designated as “the stack”.

The software stack divides the software activity into layers that
are complementary to each other.

Most companies specialize on one or few layers and rely on
other companies to offer the complementary components.

Each of these components is layered above the other, and
communicates through more or less standard interfaces, with
closer layers being more related to each other than layers that
are further apart on the stack.
5
Stack
The Horizontal Computer Industry (circa 2000)
The Vertical Computer Industry (circa 1970)
Services
IT Consulting
Sales &
Retail Stores
distribution
Systems
integration
Outsourcing Maintenance Financing
Direct
VARs
Superstores
Others
Others
Enterprise applications
Application Personal Productivity
E-Commerce Others
software (Word, Excel, Powerpoint) (Supply chain, HR, CRM)
M iddleware
Database
(Oracle, DB2)
Operating Windows and DOS
system
Computer
platform
Basic
circuitry
IBM
DEC
NCR
Sperry
Univac
Dell
Compaq
Intel
Collaboration &
Messaging
Unix
Sun
Application&
Transaction
Servers
Linux
Apple
IBM
Motorola
Mac
HP
Others
Others
Others
RISCs
Others
Wang NEC
6
Operationalization

We use the concept of “stack” to operationalize complementarity.

Construction of an index (STACKD) that measures the distance on the
stack between two companies.

The index is the weighted sum of coefficients that represents the
distance on the stack between two different layers or industry
segments. The weights are equal to the product of the percentage of
sales of each firm in the corresponding layer.

Example of a two-layer stack Hardware/Software :
0
Coefficients: d ij  
1
if i  j
if i  j
The value of the index is equal to
STACKD  PAH PTH * 0  PAH PTS * 1  PAS PTH * 1  PAS PTS * 0
 PAH PTH * 0  PTS * 1  PAS PTH * 1  PTS * 0 
7
Stack Distance Index for a L-layer Stack:
L
L
STACKD   PAi PTj dij
i 1 j 1
where
STACKD denotes stack difference index
L is the number of layers of the stack
PAi is the percentage of sales of acquirer in layer i of the stack
PTj is the percentage of sales of target in layer j of the stack
dij is a factor that assumes different values according to the distance in the stack between
layer i and layer j
8
Event Studies Methodology
Event studies are widely used in the finance and economics literature. Most
empirical studies of M&As use the event studies methodology.
This methodology is based on the assumption that share prices are simply
the present value of expected future cash flows to shareholders and that any
changes in the company’s prospects are immediately reflected in its stock
price.
1. Computation of what would be “normal returns,” using market models.
2. Computation of Cumulative Abnormal Returns for a specific event
window (we use a three-day window surrounding the M&A announcement
date).
Abnormal Returns = Actual Returns – “Normal Returns”
9
Empirical design and data:
 Study of the effects of concentration/diversification on layers of the
stack between acquirer and targets of Mergers and Acquisitions in
complementary network systems.
 Data:
Securities Data Company (SDC): data on M&As
Compustat Industry segments: sales classified as Hardware,
Software or Services (3-layers stack)
IDC: sales classified as Hardware, Systems software,
Middleware software, Applications software and Services (5layer stack)
10
Three-layer stack
For the sample obtained from Compustat we use a simplified version of the
stack (because there are only three layers):
Concentration = PAHPTH+PASfPTSf+PASvPTSv
where PAH, PAsf and PASv are the proportion of sales of the acquirer in
hardware, software and services and PTH, PTSf and PTSv are the proportion of
sales of the target in hardware, software and services.
11
Results
The low level of significance obtained for the results suggests that the
concentration or diversification on the three-layer stack explains little
of the variation of abnormal returns in M&As in the software industry.
To test if the software stack does indeed characterize the industrial
organization of the software industry, and can be the structure of a
measure of complementarity between different software companies, we
repeat the analysis using the five-layer stack and the index STACKD.
12
Results for the five-layer stack (IDC):
•The results are still not highly significant if we use the
concentration measure
•The results are not significant if we use the index STACKD.
•The results are significant when we test for curvilinear
relationship (add the square of STACKD).
13
Conclusions

We conclude that there is an inverse curvilinear relationship between
abnormal returns and the stack distance index:
Abnormal returns are higher when both sides in the acquisition are
classified in adjacent layers of the stack and smaller when acquirer
and target are further apart on the stack or are in the same layer.

The results provide evidence that there is value in M&As between
complementary components of network systems.

The results empirically validate the software stack.
14
Results

The value of a merger between software companies depends a
great deal on how easy it is to technical integrate the products of
both companies.

Very often the outcome of mergers between similar software
companies is not very successful because these companies
have problems with the technical integration of the software
products.

When companies produce in different layers of the stack,
products are already working as complementary as components
of a network system, through a common platform, and technical
integration is not an uncertainty.

M&As, allowing firms to hold equity stakes in complementary
companies, may lead to the realization of value from the
internalization of complementary network externalities
15
Extensions:

Repeat the analysis with a larger sample from IDC (five-layer
stack)

Repeat the analysis for a sample of alliances
16
Results for the three-layer stack:
Variable
Combined
Acquirer
Target
Intercept
0.1273**
0.0574
0.0352
(2.3965)
(1.1992)
*
(0.3812)
*
Concentration in Layers of Stack -0.0292
-0.0279
-0.0417
(-1.7956) (-1.7068) (-0.7347)
Payment with Cash
Acquirer Equity Value
Relative Size Target/Acquirer
0.0609***
0.063***
0.2599***
(3.7574)
(3.8654)
(4.757)
***
**
-0.0093
-0.0064
(-2.7722)
(-2.0265)
0.0355
**
(2.436)
0.002**
(2.1062)
Percentage of Target Acquired
R^2
F-statistic
N
***
Statistically significant at the 1% level.
**
Statistically significant at the 5% level.
*
Statistically significant at the 10% level.
0.1045
8.149
193
0.0915
6.345
193
0.1184
8.283
193
17
Results for the five-layer stack:
The results are still not highly significant if we use the
concentration measure:
Variable
Intercept
Model 2
0.0092
(0.6545)
Size Target/Acquirer
0.1168***
(2.8548)
Same layer / Concentration
-0.0429*
(-1.7804)
R^2
F-statistic
N
***
**
*
0.1943
5.064
45
Statistically significant at the 1% level.
Statistically significant at the 5% level.
Statistically significant at the 10% level.
18
Results for the five-layer stack:
The results are not significant if we use the index STACKD.
Variable
Intercept
Model 4
0.0056
(0.1997)
Size Target/Acquirer
0.1027**
(2.3797)
STACKD2
-0.0044
(-0.4039)
R^2
F-statistic
N
0.1368
3.329
45
***
**
*
Statistically significant at the 1% level.
Statistically significant at the 5% level.
Statistically significant at the 10% level.
19
Results for the five-layer stack:
The results are significant when we add the square of
STACKD.
Variable
Intercept
Model 5
-0.0935*
(-1.8479)
Size Target/Acquirer
0.1067*
(2.5923)
STACKD2
0.0919**
(2.1312)
STACKD2^2
-0.0195**
(-2.3019)
R^2
F-statistic
N
***
**
*
0.2356
4.213
45
Statistically significant at the 1% level.
Statistically significant at the 5% level.
Statistically significant at the 10% level.
20
Distribution of SIC codes in the three-layer stack
Hardware:
3571 – Electronic Computers
3572 – Computer Storage Devices
3575 – Computer Terminals
3577 – Computer Peripheral Equipment, Not Elsewhere Classified
3661 – Telephone and Telegraph Apparatus
3663 – Radio and Television Broadcasting and Communications Equipment
3669 – Communications Equipment, Not Elsewhere Classified
3674 – Semiconductors and Related Devices
4812 – Radiotelephone Communications
4813 – Telephone Communications, Except Radiotelephone
4822 – Telegraph and Other Related Devices
4841 – Cable and Other Pay Television Services
4899 – Communications Services, Not Elsewhere Classified
Software:
7371 – Computer Programming
7372 – Prepackage Software
Services:
7373 – Computer Integrated Systems Design
7374 – Computer Processing and Data Preparation and Processing Services
7375 – Information Retrieval Services
7376 – Computer Facilities Management Services
7377 – Computer Rental and Leasing
7378 – Computer Maintenance and Repair
7379 – Computer Related Services, Not elsewhere Classified
21