Title: An Empirical Study of the Causal Relationship Between IT Investment and Firm Performance
1An Empirical Study of the Causal Relationship
Between IT Investment and Firm Performance
- Hu, Q. and Plant, R. IRMJ, 14(3), 2001, pp. 15-26
2Outline
- Introduction
- Research background
- Research model and hypotheses
- Data and method
- Results
- Discussions
- Conclusions
3Introduction
- Productivity paradox
- What value does IT add to an organization?
- The literature in 80s and 90s contend that IT
can - provide competitive advantages
- add value
- improves operational performance
- reduces costs
- increases decision quality, and
- enhances service innovation and differentiation.
4Introduction (cont.)
- The underlying theory
- Effective use of IT ? improvement in production,
revenue, and profit - Several empirical studies support the arguments
- Brynjolfsson and Hirt (1996)
5Introduction (cont.)
- However, not all studies of industry and firm
level financial data have shown positive causal
relationship between IT investment and improved
firm performance. - Loveman (1994) found that IT investment has a
negative output elasticity. - The figure implies that the marginal dollar would
have been better spent on other categories of
capital investments.
6Introduction (cont.)
- Closer examinations of these studies revealed a
flaw in the methodologies - The impact of IT on firm performance was tested
using the IT capital data and the performance
data of the same period. - Under such circumstances, the correlation between
IT capital variables and the firm performance
variables has no inherent implication of a casual
relationship, no matter how this correlation is
established. - Why?
7Introduction (cont.)
- In this study, the authors investigate the impact
of IT investment on firm productivity and
performance using well accepted casual models
based on firm level financial data.
8Introduction (cont.)
- It is unlikely that using concurrent IT and firm
performance data would yield conclusive causal
relationship between the two. - Arguments
- IT investment ? performance
- Performance ? IT investment
9Research background
- MIS literature contend the value of IT.
- However, it is difficult to discern the added
value from business financial data. - The main reason is the inability of organizations
to track the return of IT investment when such
investment may cross many business processes and
activities.
10Research background (cont.)
- It is difficult for IS managers to convince CEO
to invest in IT projects when other capital
spending opportunities exist. - We need empirical evidences.
11Research background (cont.)
- Measuring IT effectiveness is always the top one
issue in IM domain. - Management pressure want to scrutinize IT
investment. - Are we sure that there is a payback on IT
investment ? - The necessity to understand IT investment
12Research background (cont.)
- Previous studies
- Alpar and Kim (1990)
- IT investment ? financial performance
- Subjects commercial banks
- Mixed results
- IT investment is negatively correlated with cost
- The relationship between the IT expense ratio and
the ROE was insignificant in six out of the eight
years studied.
13Research background (cont.)
- Mahmood and Mann (1993)
- Use Pearson correlation and Canonical
correlations - Test 6 organization performance variables and 6
IT investment variables - Subjects Computerworld Premier 100 companies
- Mixed results
14Research background (cont.)
- A summary of the major studies reviewed above is
presented in Table 1. - Overall, the literature on the IT impact on firm
performance has been overwhelmingly positive. - Some studies asserted the causality.
- Some used the correlation method.
- Few used explicit casual models.
15Research background (cont.)
- Correlation ? related
- Correlation ? causality
- It is possible
- IT investment ?? firm performance
- The assumption of Hirt and Brynjolfsson (1996)
- The correlation-based models will not discover
the true relationship between IT investment and
firm performance.
16Research background (cont.)
- Another flaw in the previous studies is using the
same time periods. - Casual relationships between two factors inferred
from concurrent data assume instantaneous
causality between the two factors. - The lagged effect of IT investment
- Osterman (1986), Brynjolfsson (1993), and Loveman
(1994)
17Research background (cont.)
- Two study objectives
- Determine whether there is a causal relationship
between IT investment and firm performance with
explicit causal modeling techniques - Determine the direction of the causal relationship
18Research model and hypotheses
- Correlation does not necessarily imply causation.
- If X causes Y, three conditions must hold.
- Time precedence
- Relationship
- Nonspuriousness
- For a relationship between X and Y to be
nonspuriousness, there must not be a Z that
causes both X and Y such that the relationship
between X and Y vanishes once Z is controlled.
19Research model and hypotheses (cont.)
- We can not use concurrent IT data and performance
data with correlation analysis.
20Research model and hypotheses (cont.)
- Porter and Millar (1985) asserted the three most
important benefits from IT in a firm - Reducing costs
- Enhancing differentiation
- Changing competitive scope
- In any of the cases or as a combined result, the
net effect of IT investment should be the
increased productivity and better financial
performance.
21Research model and hypotheses (cont.)
- IT benefits come not form replacing old computers
with new ones, in which the effect of investment
can be realized immediately, but from
organizational and procedural changes enabled by
IT. - The effect of such changes may take years to
realize.
22Research model and hypotheses (cont.)
- Lagged effect
- IT projects usually take years to implement.
- Organization adaptation
- Employees need time to be trained and re-skilled.
- Finally, customers and the market are the last of
these time-delayed chain reactions to respond
which ultimately determines the firm performance.
23Research model and hypotheses (cont.)
Previous IT investments
Annual Sales Growth
Operating Cost reduction
Profitability improvement
Present IT investments
Productivity improvement
24Research model and hypotheses (cont.)
- H1a The increase in IT investment per employee
by a firm in the preceding years may contribute
to the reduction of operating cost per employee
of the firm in the subsequent year.
25Research model and hypotheses (cont.)
- Figure 1 shows the research model.
- The solid arrow lines (the study)
- The dashed arrow lines (previous studies)
- It is reasonable to argue that the opposite
causal relationships exist between IT investment
and firm performance.
26Research model and hypotheses (cont.)
- H1b The reduction of operating cost per employee
by a firm in the preceding years may contribute
to the increase in IT investment per employee of
the firm in the subsequent year.
27Research model and hypotheses (cont.)
- contribute to replaces cause
- Interfering factors exist
- Operational, technological, and economic factors
- The authors have no control over these factors.
28Data and method
- It is important to obtain reliable company
IT-related data. - However, it is difficult.
- Most companies regard these data as private and
competitive information.
29Data and method (cont.)
- Important sources
- ComputerWorld database
- InformationWeek database
- Compustat database
30Data and method (cont.)
- To test the hypotheses, we need data for at least
4 consecutive years. - Preceding years,
- Present year, and
- Subsequent year(s)
31Data and method (cont.)
- Constraints three separate data sets
- Figure 2 shows the sample characteristics.
- Annual revenue
- Industry
- Annual IT spending
- Size.
- Method
- Granger causal model
32Data and method (cont.)
- Let Xt and Yt be two time series data, the
general causal model can be written - Xt b0 Yt ? aj xt-j ? bj Yt-j e
- Yt c0 Xt ? cj xt-j ? dj Yt-j ?
- If some bj is not zero, Y causes X
- If some cj is not zero, X causes Y
- If both of these event occurs, there is a
feedback relationship between X and Y. - If b0 is not zero, the instantaneous causality is
occurring and Yt causes Xt - If c0 is not zero, the instantaneous causality is
occurring and Yt causes Xt
33Data and method (cont.)
- Substituting X and Y in the casual model with
firm IT data and performance data, we can derive
a set of models for testing the research
hypotheses. - To minimizing the impact of firm size, we used
per employee metrics.
34Data and method (cont.)
- IT investments
- Equation 2
- Operating costs
- Equation 3
- Sale growth
- Equation 4
- Productivity
- Equation 5
- ProfitabilityROA, ROE
- Equations 6, 7
35Results
- Consider the inflation factor
- We inflated the financial figures of the
preceding years to the real dollar values of the
subsequent year (t) based on the annual
percentage change of implicit price deflator of
the Gross Domestic Product.
36Results (cont.)
- Because we are using the year-to-year changes as
variables, the upper limit (n) for subscript j in
all models is two ( j 2). - Use SAS software
- The results are presented in Tables 3 to 7.
- These results are summarized in Tables 8 and 9.
37Discussion
- Table 8 shows
- No convincing evidence that IT investments in the
preceding years have made any significant
contribution to the subsequent changes in any of
the four categories of firm performance measures
operating cost, productivity, sales growth, and
profitability. - The only noticeable significant b parameter is
the one for the effect of IT investment on the
ROA in the 1990-1993 data set.
38Discussion (cont.)
- However, given the overall non-significant tone
of the results, this one case of significance is
not enough to be considered as convincing
evidence to conclude that IT investment has a
positive impact on firm profitability.
39Discussion (cont.)
- Table 9 shows
- There is clear evidence to support the hypotheses
that firms budget their IT investment based on
the financial performance of preceding years,
especially the sales growth. - The faster the sale growth was achieved, the more
money was allocated for IT investment.
40Discussion (cont.)
- R2 (Tables 3 7)
- When IT investment is used as the effect and the
measures of financial performance as the causes,
most F are significant and R2adj are at decent
levels. - When the measures of financial performance are
used as the effect and IT investment as the
cause, most F are insignificant and R2adj are
very small.
41Discussion (cont.)
- We can not find the instantaneous causality
between IT investment and firm performance. - instantaneous causality b0 c0 are
significantly different from zero. - We can not find the figures in Tables 3 -7.
- We cast serious doubt on the research methodology
that uses concurrent data for testing causal
relationship between IT investment and firm
performance.
42Discussion (cont.)
- Constraints
- We do not consider the effects of industry
differences and IT maturity levels
43Conclusions
- We have shown the hypothesized positive casual
relationship between IT investment and firm
performance cannot be established at acceptable
statistical significant levels
44Conclusions (cont.)
- On the other hand, there is clear evidence that
firms had budgeted IT investment based on the
financial performance of the preceding years,
especially the growth rate of annual sales.
45Conclusions (cont.)
- Implications
- IT budget allocation
- Overspending in IT by firms may be another
complicating factor. - It has become so easy to spend a lot of money on
hardware, software, and maintenance -- and not
necessarily see any return - IT asset management
46Conclusions (cont.)
- Measure is a big problem.
- Economic value of IT
- Present measure ROE, ROA
- Barua et al. (1997) advocated the use of
intermediate variables to study the impact of IT
since they reflect the direct impact of IT
investment. - Capacity utilization, inventory turnover
47Conclusions (cont.)
- Brynjolfsson (1996) suggested
- If IT investment ? producers performance can
not be shown, we can use the surplus concept. - Consumer surplus
- Debate
- Whether it is necessary to measure the value of
IT investment - CEO care profitability!
48Conclusions (cont.)
- It seems that we have raised more questions than
provided answers in this study. - How to measure IT value?