In particular, as proposed in Aruoba, Diebold, Nalewaik, Schorfheide, and Song (ADNSS) , one can view both GDP_E and GDP_I as noisy indicators of underlying latent true GDP, which can then be extracted using optimal filtering methods.
Backtracking a bit: GDE and GDI are two sides of the same ledger and so give out the same information. In other words, although over short periods of time they can diverge, the divergence cannot be persistent or systematic. The chart shows that´s true for NGDE (NGDP) and NGDI. The recession of 1981/82 and the great recession of 2008/09 are circled.
What ADNSS do is filter out at least some of the ‘noise’ in both series to come up with a less bouncy measure.
They provide a series for the annualized growth rate of the GDP plus and the charts below compare that measure with annualized growth for both NGDP and NGDI for three selected periods of ‘Greatness”: The Great Inflation (1965-79), the Great Moderation (1987/07) and the Great Recession (2008/13).
Read the whole article here.