Kwiatkowski-Phillips-Schmidt-Shin test for stationarity
kpss varname [if exp] [in range] [, maxlag(#) notrend qs auto]
kpss is for use with time-series data. varname may contain time-series operators; help varlist. You must tsset your data before using kpss; see help tsset. kpss supports the by prefix, which may be used to operate on each time series in a panel. Alternatively, the if qualifier may be used to specify a single time series in a panel.
kpss performs the Kwiatkowski, Phillips, Schmidt, Shin (KPSS, 1992) test for stationarity of a time series. This test differs from those "unit root" tests in common use (such as dfuller, pperron and dfgls) by having a null hypothesis of stationarity. The test may be conducted under the null of either trend stationarity (the default) or level stationarity. Inference from this test is complementary to that derived from those based on the Dickey-Fuller distribution (such as dfuller, pperron and dfgls). The KPSS test is often used in conjunction with those tests to investigate the possibility that a series is fractionally integrated (that is, neither I(1) nor I(0)): see Lee and Schmidt (1996). As such, it is complementary to gphudak, modlpr and roblpr.
The test's denominator--an estimate of the long-run variance of the timeseries, computed from the empirical autocorrelation function--may be calculated using either the Bartlett kernel, as employed by KPSS, or the Quadratic Spectral kernel. Andrews (1991) and Newey and West (1994) indicate that the latter kernel yields more accurate estimates of sigma-squared than other kernels in finite samples." (Hobijn et al., 1998, p.6)
The maximum lag order for the test is by default calculated from the sample size using a rule provided by Schwert (1989) using c=12 and d=4 in his terminology. The maximum lag order may also be provided with the maxlag option, and may be zero. If the maximum lag order is at least one, the test is performed for each lag, with the sample size held constant over lags at the maximum available sample.
Alternatively, the maximum lag order (bandwidth) may be derived from an automatic bandwidth selection routine, rendering it unnecessary to evaluate a range of test statistics for various lags. Hobijn et al. (1998) found that the combination of the automatic bandwidth selection option and the Quadratic Spectral kernel yielded the best small sample test performance in Monte Carlo simulations.
Approximate critical values for the KPSS test are taken from KPSS, 1992.
The KPSS test statistic for each lag is placed in the return array.
maxlag(#) specifies the maximum lag order to be used in calculating the test. If omitted, the maximum lag order is calculated as described above.
notrend indicates that level stationarity, rather than trend stationarity, is the null hypothesis.
qs specifies that the autocovariance function is to be weighted by the Quadratic Spectral kernel, rather than the Bartlett kernel.
auto specifies that the automatic bandwidth selection procedure proposed by Newey and West (1994) as described by Hobijn et al. (1998, p.7) is used to determine maxlag. In that case, a single value of the test statistic is produced, at the optimal bandwidth.
. use http://fmwww.bc.edu/ec-p/data/macro/nelsonplosser.dta . kpss lrgnp
. kpss D.lrgnp, maxlag(8) notrend . kpss lrgnp if tin(1910,1970) . kpss lrgnp, qs auto
Christopher F. Baum, Boston College, USA email@example.com
References Andrews, D.W.K. Heteroskedasticity and Autocorrelation Consistent Covariance Matrix Estimation. Econometrica, 59, 1991, 817-858.
Hobijn, Bart, Franses, Philip Hans, and Marius Ooms. 1998. Generalizations of the KPSS-test for Stationarity. Econometric Institute Report 9802/A, Econometric Institute, Erasmus University Rotterdam. http://www.eur.nl/few/ei/papers
Kwiatkowski, D., Phillips, P.C.B., Schmidt, P. and Y. Shin. Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root? Journal of Econometrics, 54, 1992, 159-178.
Lee, D. and P. Schmidt. On the power of the KPSS test of stationarity against fractionally-integrated alternatives. Journal of Econometrics, 73, 1996, 285-302.
Newey, W.K. and K.D. West. Automatic Lag Selection in Covariance Matrix Estimation. Review of Economic Studies, 61, 1994, 631-653.
Schwert, G.W. Tests for Unit Roots: A Monte Carlo Investigation. Journal of Business and Economic Statistics, 7, 1989, 147-160.
A version of this code written in the RATS programming language by John Barkoulas served as a guide for the development of the Stata code. Thanks to Richard Sperling for suggesting its validation against the Nelson-Plosser data (KPSS, 1992, Table 5).
On-line: dfuller, pperron, time, tsset, dfgls, gphudak (if installed), modlpr (if installed), roblpr (if installed)