• II.

Fixed effects within-group model

• This model also pools all 90 observations but treats each airline
separately.

• It involves expressing each variable as a deviation from its mean

value within each airline and then estimating an ordinary least
squares (OLS) regression on these mean-corrected or ”de-meaned”
values.

• This approach controls for time-invariant unobserved factors and

captures the within-group variation.
• One way to estimate a pooled regression is to eliminate the
fixed effect, β1i , by expressing the values of the dependent
and explanatory variables for each airline as deviations from
their respective mean values.
• The Within-Group Estimator removes αi by demeaning the
data. This is done by subtracting the entity-specific mean from
each variable:
 After, eliminating αi, the model becomes:

 Where the de-meaned value of each variable can be generate by subtracting the
mean value of each variable from the respective value for each airline.

 The table presents the results from a fixed-effects panel data model, specifically
a within-group model, aimed at estimating the cost function based on annual
data for six airlines over a period of 15 years.
• The model utilizes de-meaned variables to capture the
variations in costs, output, fuel prices, and load factors after
removing the average values for each airline-year
combination. The estimated coefficients reveal the
relationships between these de-meaned variables and the cost
function.
c Coef. Std. Err. t P>t [95% Conf. Interval]

DMQ 3319023 171354.1 19.37 0.000 2978083 3659964

DMPF .7730708 .097319 7.94 0.000 .5794365 .9667052
lDMLF -3797368 613773.1 -6.19 0.000 -5018584 -2576152

sigma_u 748483.04
sigma_e 210422.77
rho .92675367 (fraction of variance to u_i)
due

R-squared =0.929366
Adjusted R-squared = 0.927743
• The estimated coefficients reveal the relationships between
these de-meaned variables and the cost function.
• For de-meaned output (DMQ), the coefficient is estimated to
be 03319023, and a t-statistic of 19.37.
• Additionally, the coefficient for de-meaned fuel price (DMPF)
is 7730708, with a standard and a t-statistic of 7.94..
• Finally, for de-meaned values of load factor(DMLF), the
coefficient is estimated to be 3797368, and a t-statistic of -6.19
.
C. Random effects model (REM)
 In contrast to the LSDV model, the REM assumes that the
intercept values are random draws from a larger population of
airlines.
 It allows for both time-invariant and time-varying unobserved
heterogeneity, capturing the individual-specific effects and the
common effects shared across airlines.
 If the dummy variables indeed indicate a lack of
understanding regarding the (true) model, then why not
incorporate this uncertainty into the error term?
• This is precisely the approach suggested by
the proponents of the so-called error
components model (ECM) or random effects
model (REM), which we will now illustrate
with our airline cost function.
The basic idea is to start with the equation:
• = ++++equ(4.6)
• Instead of treating as fixed, we assume that it
is a random variable with a mean value of (no
subscript i here).The intercept value for an
individual company can be expressed as:
• =+ Equ(4.7)
• where is a random error term with a mean
value of zero and a variance of
• What we are essentially saying is that the six firms included in
our sample are a drawing from a much larger universe of such
companies and that they have a common mean value for the
intercept (= ).
• The individual differences in the intercept values of each
company are reflected in the error term .
• Substituting Eq. (4.7) into Eq. (4.6), we obtain:
• = ++++
• =++++
• Where,
• The composite error term wit consists of two components: εi ,
which is the cross-section, or individual-specific, error
component, and uit , which is the combined time series and
cross-section error component and is sometimes called the
idiosyncratic term because it varies over cross-section (i.e.,
subject) as well as time.
• The error components model (ECM) is so named because the
composite error term consists of two (or more) error
components.
• The usual assumptions made by the ECM are
that:
• In FEM each cross-sectional unit has its own (fixed) intercept
value, in all N such values for N cross-sectional units.
• In ECM, on the other hand, the (common) intercept represents
the mean value of all the (cross-sectional) intercepts and the
error component εi represents the (random) deviation of
individual intercept from this mean value.
• Keep in mind, however, that εi is not directly observable; it is
what is known as an unobservable, or latent, variable.
• As a result of the assumptions stated in the
above equation, it follows:
• It is also very important to note that wit is not correlated with
any of the explanatory variables included in the model.
• Since uit is a component of wit , it is possible that the latter is
correlated with the explanatory variables.
• If that is indeed the case, the ECM will result in inconsistent
estimation of the regression coefficients.
 Regression Results
c Coef. Std. Err. z P>z [95% Conf. Interval]

q 2288588 109493.7 20.90 0.000 2073984 2503192

pf 1.123591 .1034406 10.86 0.000 .9208515 1.326331
lf -3084994 725679.8 -4.25 0.000 -4507301 -1662688
_cons 1074293 377468 2.85 0.004 334469.4 1814117

sigma_u 107411.2
sigma_e 210422.77
of variance
rho .20670403 (fraction to u_i)
due
Prob > chi2 = 0.0000
Wald chi2(3) = 883.50

sigma_u = 107411.2
sigma _e = 210422.77
 Hausman Test
• When comparing the results of the fixed-effects and random-
effects regressions, we often observe significant differences
between them. This raises the important question of which
results are more reliable and which model should be chosen.
• The Hausman test is a statistical test used to determine
whether the fixed-effects model (FEM) or the random-effects
model (REM) is more appropriate for a given dataset.
• It helps us assess the reliability of the results obtained from
these two models.
• The Hausman test statistic, developed by Hausman, has an
asymptotic distribution that allows us to assess whether the
differences between the FEM and REM estimators are
statistically significant.
• If the test statistic shows that the null hypothesis cannot be
rejected, it suggests that the random effect model is
appropriate.
• However, if the test statistic indicates a significant difference
between the two estimators, it implies that the fixed effect
model is more appropriate than the random effect model.
• If the p-value is small (typically < 0.05): Reject the null
hypothesis and conclude that the Fixed Effects (FE) model is
preferred. The individual effects are likely correlated with the
explanatory variables, and using RE would lead to inconsistent
estimates.
• If the p-value is large (typically > 0.05): Fail to reject the
null hypothesis and conclude that the Random Effects (RE)
model is appropriate.
• The individual effects are uncorrelated with the explanatory
variables, and RE is more efficient.