Title: Campus Presentation at National Taiwan University
1Campus Presentation at National Taiwan University
- Wesley Shu
- Assistant Professor
- San Diego State University
2Short Biography
- BA in Economics, National Taiwan University
- MBA in Finance Decision Sciences, Indiana
University - Ph.D. in MIS, University of Arizona
3IT Productivity and Productive Efficiency in
Taiwan
4What is Productivity?
- The amount of output produced given an input
- Output/Input
- What if multiple input?
5A Cobb-Douglas Function
6Other Functional Forms
7Productivity and Productive Efficiency
- Productivity is to measure how much business
value an input factor can contribute to. - Productive efficiency is to measure the gap
between observed and optimal values of output and
input.
8Three Types of Inefficiency
- Technical inefficiency
- Allocative inefficiency
- Scale inefficiency
9Technical Inefficiency
- The gap between the observed output and the
production frontier under the current technology
10Technical Inefficiency, continued
x2
Production frontier
A
B
x1
11Allocative Inefficiency
- A firm chooses the input ratio when the marginal
ratio input price ratio to minimize its total
cost. When they are not equal,
12Scale Inefficiency
- A firm chooses its production level when the
marginal cost output price. If not, then
13Characteristics of Previous Studies
- Measuring single deterministic production
function - Not incorporating some basic business assumptions
14Deterministic Approach
- Deterministic approach assumes all deviations
except the error terms are under management
control - It in fact uses observed data to construct the
production frontier (optimal output level.)
15Not Imbedding the Basic Business Assumption
- Firms want to either maximize their profits or
minimize their costs - So, they will decide the output and input
quantities based on the price information - This price information and firms decision
behavior are not captured in a single production
function approach, but in the error terms. - So, there is bias because the explanatory
variables are correlated with the error terms.
16Not Imbedding the Basic Business
Assumptioncontinued
- Hal Varian, Microeconomic Analysis, 3rd Edition
- If the managers observe these effects (of price
changes,) then they will certainly take that
information into account when they determine
their optimal choice of inputs. Thus, the
right-hand variables (of a production function)
will not be statistically independent from the
error term.
17Our Model - formalize the business assumption
Profit maximization model with inefficiency
measurement
18Our Model, continued
Endogenous variable xi Exogenous variable p, wi
19Our Model, continued
- Intrilligator, Bodkin, and Hsiao
- Estimating the complete system is generally
superior to estimating only the first equation
(the production function) from both economic and
econometric standpoints. - From an economic standpoint, estimating the
complete system expresses the assumption that the
data reflect both the behavior of the decision
maker (the firm) and the technology, while the
first equation (the production function) reflects
only the technology. - From an econometric standpoint, the estimators of
only the first equation involves simultaneous
equations bias, so the estimators will be biased
and inconsistent.
20Data Requirement
- Assets, output price, Employment Compensation are
publicly available. - IT Employment Compensation
- IT Spending, including hardware, software,
maintenance, and training - Prices (Price deflators)
21Our Production Function
22Data Source
- Year 2000 - 2002
- Survey of more than 300 companies,
- 187 with valid data (all three years)
- A variety of industries
23Data Requirement IT Capital
- From survey
- The survey is IT Spending. Need to convert
flow into stock. - Companies may know spending but not stock or
asset. - Since we only have 3-year data, we assume IT life
cycle is 2 years.
24Data Requirement IT Capital
- Rate of depreciation, ex., in a year, ½
of IT to be obsolete.
25Data Requirement IT price
- Rental price very complicated formula
- Our research survey
26Finding, Productivity
27Findings - Inefficiency
- Technical -0.1350
- Allocative uNIT -0.7989 decrease non-IT
- uLIT 0.6138
- uNLIT -0.2314
- Scale 0.2423 over produce
28Findings - Overall Percentage Loss
29Future Direction
- After March when 2003 data available complete
the research - Add panel data consideration into the model
30Analysis of Panel Data
- Cross section and time series
- With consideration of stochastic form or not
Y
Company C
Company B
Company A
I
31Future Direction continued
- Put into consideration the company size and
industry difference - Relax constraints - CES
- Measure input substitution effect translog
function