Firm Size and Information Technology Investment:

1
Firm Size and Information Technology Investment
Beyond Simple Averages
Tianyi Jiang Leonard N. Stern School of
Business New York University December 16, 2003
2
Motivation
Effectiveness of information technology (IT)
investments Specifically How does IT impact
firm sizes and firm boundaries
3
ITs Theoretical Impact on Firm
IT decreases decision cost, agency cost,
coordination costs between across firms
- Gurbaxani et al. 1991 could make firms
smaller - Malone et al. 1987 lead to
outsourcing from fewer suppliers - Bakos et
al. 1993
4
Empirical Evidence on the impact of IT
IT investment is negatively correlated with
firm size across all industries.
Brynjolfsson, et al. 1994 is negatively
correlated with vertical integration is weakly
positively correlated with diversification
Hitt, 1999
5
Research Questions

• In the context of new NAICS classifications, are
  IT
• investments negatively correlated with firm size
  across
• all industries?
• In measuring impact of IT investments at the
  industry
• level, is average employees per firm a
  good measure?

6
NAICS Industries IT investment ratio in 1992
7
Regression on 1992 COMPUSTAT Data
13/16 industries significant at 99 level
  Key Significant at 90 level
Significant at 95 level Significant at 99
level
8
Problems with simple firm averages

• Observations
• Large numbers of small firms can bring down
  average
• firm sizes even if the bigger firms got bigger
• example firms sizes 1,1,1,1,100,100
• average size 34
• Most entry exit has relatively little effect
  on the largest
• firms in the industry - Sutton 1997

9
Problems with median firm sizes
Median
Median
1 total Emp
99 total Emp
.8 total Sales
99.2 total Sales
1992 Professional Services Employee Histogram
1992 Professional Services Sales Histogram
10
Employee weighted firm sizes

• Emphasize the size of larger firms to minimize
  the effects
• of entry exit - Kumar et al. 2001
• Weighted Average Number of Employees
• total number of employees in a bin
• total number of employees in the
  sector
• total number of firms in a bin

Kumar, K., Rajan, R., Zingales, L. What
Determines Firm Size? Working Paper, The
University of Chicago Graduate School of
Business, 2001.
11
Employee size calculation example
Example firms sizes 1,1,1,1,100,100
average size 34 Employee weighted average
2 bins 1,1,1,1 and 100, 100
weighted average (4/204)(4/4)
(200/204)(200/2)
98.05882
12
Automated bin partition
Recursive Minimum Entropy Partitioning Fayyad
et al. 1993 Entropy A measure of homogeneity of
values Mitchell 1997 Example 2 distinct
values, i,j, i?j S be a bin of
firms with i or j employees then
Pi percentage of firms with i employees
13
Recursive Minimum Entropy Partitioning
Let S original bin A set of newly split
bins Gain (S,A) Entropy(S)-EEntropy(A) I
dea Recursively split data into smaller bins
with nearly homogenous values until gain
lt threshold
14
Recursive Minimum Entropy Partitioning (cont.)
Recursive Splits
15
Sales weighted firm sizes

• Alternatively, we could emphasize firms with
  higher
• proportion of sales to minimize the effects
• of entry exit
• Sales Weighted Employees Sizes
• total number of employees in a bin
• total number of firms in a bin
• total amount of sales in a bin
• total amount of sales in a sector

16
Firm size measures across NAICS industries with
low IT investment ratio
17
Firm size measures across NAICS industries with
low IT investment ratio (cont.)
18
Firm size measures across NAICS industries with
medium IT investment ratio
19
Firm size measures across NAICS industries with
high IT investment ratio (cont.)
20
Regression Model
natural log of 3 different employee
measures in year t
natural log of IT investment ratio per industry
in year t natural log of net
sales per industry per year 17
industry dummy variables i.i.d.
error term with zero mean
21
Data Methodology

• Computed industry level employee measure net
  sales
• via COMPUSTAT data from 1982 to 2001
• (443,507 records)
• Extracted IT investment ratio from BEA (Bureau
  of
• Economic Analysis) Input-Output use tables for
  the
• benchmark years of 1982, 1987, 1992, 1997
• (Required many to many mappings of NAICS to SIC
• and SIC to IO codes)
• Interpolated IT investment ratios for other
  years

22
Regression Results Across 6 NAICS Industries
23
Regression Result Professional Services
24
Research Limitations

• Need yearly IT investment data across all
  industries
• Tried Brookings panel data, replicated previous
  results
• across industries, but lacked the data for
• Professional Services

25
Summary

• Technical Research Contributions
• Apply recursive minimum entropy methods to the
• empirical economics domain
• Economic Research Contributions
• Utilize weighted average employee sizes to
  replicate
• previous studies on IT investments and firm
  sizes
• Found varying patterns of evolving firm sizes
  across
• industries with different IT investment
  ratios

26
Thank You!
Special thanks to Ramesh Sankaranarayan Shinkyu
Yang