Page 125 - Invited Paper Session (IPS) - Volume 2
P. 125

IPS 188 G. P. Samanta
            3.3  Correlation
                The tracking ability of GMPrice or GMInfl is examined simply by correlation
            coefficient between GMPrice/GMInfl and price index either in their original or
            stationary-transformed  forms.  Significance  of  these  correlation  coefficients
            would  establish  the  tracking  or  nowcasting ability  of  Google  search  index.
            Table 4 presents correlation coefficients for (a) different pairs of annual rate of
            inflation or change in  CPIC, CPI-U, GMPrice and GMInfl; (b) monthly change
            (i.e.  first  difference)  of  the  annual  inflation  or  growth  rates; and  (c)  similar
            results for stationary-transformed series for lnCPI-C, lnCPI-U, lnGMPrice, and
            lnGMInfl.  As  seen,  the  annual  percentage  change  in  GMInfl  is  strongly
            correlated  with  annual  inflation  rates  based  on  both  CPI-C  and  CPI-U,
            indicating that Google search index GMInfl is useful in tracking annual inflation
            rates. Further, positive sign of correlation coefficients indicates the pairs of
            variables usually movement in same direction.

                 Table 4: Correlation Coefficient between Different Pairs of Variables
             Variable Pair     Correlation       Variable Pair        Correlation
                               Coefficient                             Coefficient
             gGMPrice & gCPI-C    0.1805   (0.1212)   gGMInfl & gCPI-C    0.3591      (0.0016)
             gGMPrice & gCPI-U    0.1641   (0.1595)   gGMInfl & gCPI-U    0.3509      (0.0020)
             ∆gGMPrice & ∆gCPI-  0.0150   (0.8985)   ∆gGMInfl & ∆gCPI-C         0.1608     (0.1704)
             C
             ∆gGMPrice & ∆gCPI- -0.0082  (0.9447)   ∆gGMInfl & ∆gCPI-U    0.1710     (0.1450)
             U
             ∆lnGMPrice &       0.1695   (0.1186)   ∆lnGMInfl & ∆lnCPI-C   0.0373      (0.7333)
             ∆lnCPI-C
             ∆lnGMPrice &       0.1401   (0.1981)   ∆lnGMInfl & ∆lnCPI-U   -0.0280     (0.7977)
             ∆lnCPI-U
             e2lnGMPrice &      0.3977   (0.0001)   e2lnGMInfl & e2lnCPI-C  0.2557    (0.0168)
             e2lnCPI-C
             e2lnGMPrice &      0.3701   (0.0013)   e2lnGMInfl & e2lnCPI-U  0.2913    (0.0062)
             e2lnCPI-U
            Figures within ( ) are p-values.

            4.  Discussion and Conclusion
                Prediction of inflation or quantification of market expectations on inflation
            is always very challenging. Other than model-based forecast, a conventional
            way of measuring market expectations involves conducting suitable surveys
            to capture current assessment and forward-looking outlook of the appropriate
            target  group.  However,  surveys  are  associated  with  substantial  cost  and
            resource  implications,  in  addition  to  facing  specific  conceptual  and
            operational challenges. As a potential alternative to address the issue, many
            researchers argue that interest on Google search on a relevant keyword, such

                                                               112 | I S I   W S C   2 0 1 9
   120   121   122   123   124   125   126   127   128   129   130