R/f_highermomentsIV.R
higherMomentsIV.RdFits linear models with one endogenous regressor using internal instruments built using the approach described in Lewbel A. (1997). This is a statistical technique to address the endogeneity problem where no external instrumental variables are needed. The implementation allows the incorporation of external instruments if available. An important assumption for identification is that the endogenous variable has a skewed distribution.
higherMomentsIV(formula, data, verbose = TRUE)Returns an object of classes rendo.ivreg and ivreg, It extends the object returned from
function ivreg of package AER and slightly modifies it by adapting the call
and formula components. The summary function prints additional diagnostic information as
described in documentation for summary.ivreg.
All generic accessor functions for ivreg such as anova, hatvalues, or vcov are available.
Consider the model:
The observed data consist of Yt, Xt and Pt, while Zt, εt, and νt are unobserved. The endogeneity problem arises from the correlation of Pt with the structural error εt, since E(εν)≠0. The requirement for the structural and measurement error is to have mean zero, but no restriction is imposed on their distribution.
Let S̅ be the sample mean of a variable St and Gt=G(Xt) for any given function \(G\) that has finite third own and cross moments. Lewbel(1997) proves that the following instruments can be constructed and used with two-stage least squares to obtain consistent estimates:
Instruments in equations \(3e\) and \(3f\) can be used only when the measurement and the structural errors are symmetrically distributed. Otherwise, the use of the instruments does not require any distributional assumptions for the errors. Given that the regressors \(G(X) = X\) are included as instruments, \(G(X)\) should not be linear in \(X\) in equation \(3a\).
Let small letter denote deviation from the sample mean: si = Si-S̅. Then, using as instruments the variables presented in equations \(3\) together with \(1\) and Xt, the two-stage-least-squares estimation will provide consistent estimates for the parameters in equation \(1\) under the assumptions exposed in Lewbel(1997).
The formula argument follows a four part notation:
A two-sided formula describing the model (e.g. y ~ X1 + X2 + P), a single endogenous regressor
(e.g. P), and the exogenous variables from which the internal instrumental variables should
be build (e.g. IIV(iiv=y2)), each part separated by a single vertical bar (|).
The instrumental variables that should be built are specified as (multiple) functions, one for each
instrument. This function is IIV and uses the following arguments:
iivWhich internal instrument to build. One of g, gp, gy, yp, p2, y2 can be chosen.
gWhich function g represents in iiv.
One of x2, x3, lnx, 1/x can be chosen.
Only required if the type of internal instrument demands it.
...The exogenous regressors to build the internal instrument. If more than one is given, separate instruments are built for each. Only required if the type of internal instrument demands it.
Note that no argument to IIV is to be supplied as character but as symbols without quotation marks.
Optionally, additional external instrumental variables to also include in the instrumental variable regression can be specified. These external instruments have to be already present in the data and are provided as the fourth right-hand side part of the formula, again separated by a vertical bar.
See the example section for illustrations on how to specify the formula parameter.
Lewbel A (1997). “Constructing Instruments for Regressions with Measurement Error When No Additional Data are Available, With an Application to Patents and R&D.” Econometrica, 65(5), 1201–1213.
data("dataHigherMoments")
# P is the endogenous regressor in all examples
# 2 IVs with g*p, g=x^2, separately for each regressor X1 and X2.
hm <- higherMomentsIV(y~X1+X2+P|P|IIV(iiv=gp, g=x2, X1, X2),
data = dataHigherMoments)
#> The following internal instruments were built: IIV(iiv=gp,g=x2,X1,X2).
#> Fitting an instrumental variable regression with model y ~ X1 + X2 + P|X1 + X2 + IIV(iiv=gp,g=x2,X1,X2).
# same as above
hm <- higherMomentsIV(y~X1+X2+P|P|IIV(iiv=gp, g=x2, X1) +
IIV(iiv=gp, g=x2, X2),
data = dataHigherMoments)
#> The following internal instruments were built: IIV(iiv=gp,g=x2,X1), IIV(iiv=gp,g=x2,X2).
#> Fitting an instrumental variable regression with model y ~ X1 + X2 + P|X1 + X2 + IIV(iiv=gp,g=x2,X1) + IIV(iiv=gp,g=x2,X2).
# 3 different IVs
hm <- higherMomentsIV(y~X1+X2+P|P|IIV(iiv=y2) + IIV(iiv=yp) +
IIV(iiv=g,g=x3,X1),
data = dataHigherMoments)
#> The following internal instruments were built: IIV(iiv=y2), IIV(iiv=yp), IIV(iiv=g,g=x3,X1).
#> Fitting an instrumental variable regression with model y ~ X1 + X2 + P|X1 + X2 + IIV(iiv=y2) + IIV(iiv=yp) + IIV(iiv=g,g=x3,X1).
# use X2 as external IV
hm <- higherMomentsIV(y~X1+P|P|IIV(iiv=y2)+IIV(iiv=g,g=lnx,X1)| X2,
data = dataHigherMoments)
#> The following internal instruments were built: IIV(iiv=y2), IIV(iiv=g,g=lnx,X1).
#> Fitting an instrumental variable regression with model y ~ X1 + P|X1 + IIV(iiv=y2) + IIV(iiv=g,g=lnx,X1) + X2.
summary(hm)
#>
#> Call:
#> higherMomentsIV(formula = y ~ X1 + P | P | IIV(iiv = y2) + IIV(iiv = g,
#> g = lnx, X1) | X2, data = dataHigherMoments)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -6.49866 -1.42920 -0.01454 1.45356 10.16266
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 4.07558 0.86687 4.701 2.72e-06 ***
#> X1 4.57633 0.43273 10.576 < 2e-16 ***
#> P -0.98017 0.08896 -11.018 < 2e-16 ***
#>
#> Diagnostic tests:
#> df1 df2 statistic p-value
#> Weak instruments 3 2495 5.90 0.000522 ***
#> Wu-Hausman 1 2496 2.28 0.131143
#> Sargan 2 NA 1015.68 < 2e-16 ***
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> Residual standard error: 2.174 on 2497 degrees of freedom
#> Multiple R-Squared: 0.8378, Adjusted R-squared: 0.8377
#> Wald test: 107 on 2 and 2497 DF, p-value: < 2.2e-16
#>