Student - Dummy Variable Issue
Uploaded by
Shadman SakibStudent - Dummy Variable Issue
Uploaded by
Shadman SakibTOPIC – DUMMY VARIABLES.
1. Question for Discussion No.1
a) Which particular value/s are commonly used as dummy variable/s in a
regression?
b) Can we use dummy variables in case of multiple groups?
c) Which type of problem may arise if we do not follow dummy variable trap
properly?
d) Consider a simple model with one continuous variable (x) and one dummy
(d). What is the difference between two models if d = 0 or 1 for comparing
the alternatives?
e) Among different types of measurement scales and data set, which one/s are
best fitted for dummy variable?
f) ‘The interaction of dummy variable is closely related to that of mediation
effect’- are you agree with this statement? How can you generate
interaction variable in regression equation?
g) Distinguish among direct, indirect & total effect of mediating variable.
2. Question for Discussion No. 9.4
From annual data for 1992-1999, William Nordhaus estimated the following model to
explain the OPEC’s oil price behavior (standard errors in parentheses for 5% sig.).
𝑦̂𝑖 = 0.3 𝑥1𝑡 + 5.22𝑥2𝑡
se = (0.03) (0.50)
where y = difference between current and previous year’s price (dollars per barrel)
x1 = difference between current year’s spot price and OPEC’s price in the previous
year
x2=1 for 1994 and 0 otherwise.
a) Interpret this result in 1994 than the other years in the sample.
b) What do these results suggest about oil price behavior?
c) Based on statistical significance, analyze your findings of A and B.
3. Question for Discussion
Consider the following model
𝑦̂𝑖 = 𝛼1 + 𝛼2 𝐷𝑖 + 𝛽𝑋𝑖 + 𝜇𝑖
where Y = annual salary of a college professor
X = years of teaching experience
D = dummy for gender
Consider following three ways of defining dummy variable:
1
a. D=1 for male, 0 for female
b. D=1 for female, 2 for male
c. D=1 for female, -1 for male.
Requirement: a) Interpret the preceding regression model for each dummy
assignment. Is one method preferable to another? Justify your answer.
b) What is the impact of dummy on intercept & slope coefficient in regression?
4. Question for Discussion No. 9.16
To study the rate of growth of population in Belize over the period 1970-1992,
Mukherjee et al. estimated the following models:
Model I: ln(̂𝑃𝑜𝑝)𝑡 = 4.73 + 0.024t
t = (81.25) (54.71)
Model II: ln(̂𝑃𝑜𝑝)𝑡 = 4.77 + 0.015t - 0.075Dt + 0.011(Dtt)
t = (477.92) (34.01) (-17.03) (25.54)
Pop = population in millions
t = trend variable
Dt = 1 for observations beginning in 1978 and 0 before 1978 and
Ln stands for natural logarithm.
a) In Model I, what is the rate of growth of Belize’s population over the sample
period?
b) Are the population growth rates statistically different pre- and post-1978? How
do you know?
c) What are the cautionary issues that analysts should consider whilst
using dummy variable?
5. Questions for discussion
In his study on the labor hours spent by the FDIC (Federal Deposit Insurance
Corporation) on 91 bank examinations, R. J. Miller estimated the following function*:
ln 𝑌̂̂ = 2.41 + 0.3674 ln X1 + 0.2217 ln X2 + 0.0803 ln X3
(0.0477) (0.0628) (0.0287)
−0.1755D1 + 0.2799D2 + 0.5634D3 − 0.2572D4
(0.2905) (0.1044) (0.1657) (0.0787)
R2 = 0.766
where Y = FDIC examiner labor hours
X1 = total assets of bank
2
X2 = total number of offices in bank
X3 = ratio of classified loans to total loans for bank
D1 = 1 if management rating was “good”
D2 = 1 if management rating was “fair”
D3 = 1 if management rating was “satisfactory”
D4 = 1 if examination was conducted jointly with the state
The figures in parentheses are the estimated standard errors.
a. Based on functional form, which type of regression it is? Interpret the results of the
model.
b. Is there any problem in interpreting the dummy variables in this
model since Y is in the log form?. How would you interpret the
dummy coefficients? Please consider the suggestion made by
Halvorsen and Palmquist to provide your response.
c. Identify the difference between Logarithm and Anti-logarithm in
economic model with example.
6. Analyze the finding of the following Probability Model. Interpret ‘Firm*Country’, firm level and country level control
variable/s.
𝑃𝑟𝑜𝑏 (𝑆𝑢𝑘𝑢𝑘𝑖𝑡 = 1|𝑋𝑖𝑡 ) = 𝛼 + 𝛽′𝑍_𝑠𝑐𝑜𝑟𝑒𝑖𝑡 + ∑𝑛𝑖=1 𝛶𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑖𝑡 + 𝜀𝑖𝑡 ; Sukuk=1 if
firm issues sukuk otherwise bond issuing firms =0, controls includes a set of firm level
Name and Illustration
and country level control variables, and finally 𝑌̂𝑒𝑎𝑟𝑖𝑡 are the year dummies in the model.
Altman Z-score Altman Z-score
Variables
(Five factor-MF) (Four factor-ALL)
Model-1 Model-2
Z Scores -.112*** -.414**
(-2.65) (-2.20)
Size 1.72*** 1.407***
(3.89) (3.48)
Cash flow per share .044 .069
(0.29) (0.47)
Profit margin .0115 .008
(0.94) (0.78)
GDP per capita -.868* -.560
(-1.73) (-1.45)
Inflation -.028 -.025
(-0.62) (-0.62)
Corruption .854*** .818*
(4.73) (1.84)
Firm*Country -.008*** -.007***
(-11.87) (-12.20)
Constant -.665 -2.717
(-0.01) (-0.02)
