Notes

• Nowcast models use up-to-the minute data to constantly update our estimate of economic activity. Using this approach, we can call the GDP of an economy long before national accounts data are released.
• We first isolate a set of high-frequency variables that are good statistical proxies for various components of GDP.
• Second, an autoregressive forecasting framework is used to fill in the data not yet available. As the quarter progresses, actual data replace forecasts and the Nowcast begins to approximate GDP.
• Third, we estimate the momentum of these variables though a traditional mean-variance approach.
• Fourth, the normalized variables are converted to a composite score of quarterly activity. Optimized z score is the weighted average of the 3mma of the z score of annual growth rates of high frequency variables. To calculate optimized composite score, each variable is weighted by its coefficient estimate obtained in a regression, where GDP is the dependent variable.