The spectral classifications of the stars from spectral data have been corrected from time to time and new spectral and luminosity classes have been assigned. Identifying stars with wrong spectral and luminosity classification has been a stupendous task from the huge catalogue of stars. In this work we describe a simple statistical technique to identify stars with wrong spectral and luminosity classification. We make use of the Hipparcos catalogue which has the most accurate measurement of the distance d of the stars. A comparison is made between the absolute V magnitudes MV computed using the observed V magnitude mV and d, with the standard absolute magnitude MV0 assigned to a spectral and luminosity classification for a large number of stars (with d < 100 pc). As expected, for most of the stars the difference between MV and MV0 lies within the range ±2 mag, due to the intrinsic nature of each star ignored in this generalisation. A systematic error analysis is made of all the observable used in the computation. Therefore to identify stars which we suspect to be wrongly classified, we look for abnormal deviation in |MV - MV0 | ≥5. The location of these stars with respect to the galactic plain and interstellar extinction is also investigated to rule out effects due to variations in the interstellar extinction. From our results we see that some of the stars were indeed wrongly classified and have recently been reclassified (SIMBAD). The reclassification drastically reduces the |MV - MV0 | deviation. The other stars in the list which have not yet been reclassified need to be spectroscopically investigated and classified again.