What The Fed's Computer Model Predicts About The Future Of The US Economy

Now that our position since about 2009 that not only economic forecasting, but the economic profession in itself is just a big joke has finally gone mainstream... 

... courtesy of none other than Yellen's own San Fran Fed which yesterday finally jumped the shark when it confirmed that all economists do is "goalseek" data, only they call it "double seasonal adjustments...

... the Fed is left scratching its head: if its carbon-based predictors of the future are so embarrassingly clueless and have to resort to goalseeking gimmicks to validate their "work", who is left?

The answer: computer models. Or rather, the NY Fed's dynamic stochastic general equilibrium (DSGE) model, which was introduced by the Goldman-controlled central bank branch in September 2014

So where human economists falter (at least most of them: apparently when plugging in numbers without any bias in a model framework as the Atlanta Fed does leads to almost uncanny accuracy, but the problem is nobody in the financial arena is able to strip out their bullish bias to any and every data), why does the Fed believe a computer model will do anything better? Here is the answer (spot the circular referneces):

[T]he FRBNY DSGE model is a macroeconomic model based on modern economic theory, which characterizes the equilibrium evolution of key macroeconomic variables and identifies the underlying shocks that perturb the economy. This model is estimated using Bayesian statistical techniques, which combine prior information on model parameters with a range of data series. The DSGE model is a work in progress. We continuously strive to improve it, and augment it, so that it can provide information about a growing set of economic variables. Accordingly, the forecasts presented here are obtained using a new version of the FRBNY DSGE model discussed in September.


This version builds on the New Keynesian model with financial frictions used in Del Negro, Giannoni, and Schorfheide (2015), which has been shown to provide a reasonable explanation for the behavior of inflation in the aftermath of the Great Recession, and relatively accurate forecasts of output growth and inflation throughout recent history. Relative to the previous version of the FRBNY DSGE model, the set of observable indicators is augmented with data on consumption and investment growth, survey-based long-run inflation expectations, which provide information on the public’s perception of the central bank’s inflation objective, and the ten-year Treasury yield, in order to incorporate information about long-term rates. In addition, the model is estimated using two distinct measures of inflation—the GDP deflator and core personal consumption expenditures (PCE) inflation. Finally, the model allows for persistent shocks to both the level and the growth rate of productivity (where the latter shocks account for the possibility of secular stagnation), and uses data on the growth rate of productivity from the San Francisco Fed in order to inform these processes.

So what does this circular reasoning model (whose output steering central bank policy is based on "information on the public’s perception of the central bank’s inflation objective") say about the future:

It continues to predict a gradual recovery in economic activity ...

No surprise there: once wrong, always wrong. And when really wrong, just double, triple, quadruple and so on seasonally-adjusted the data until you get what you want....

... with a progressive but slow return of inflation toward the Federal Open Market Committee’s (FOMC) long-run target of 2 percent. This forecast remains surrounded by significant uncertainty. Please note that the DSGE model forecasts are not the official New York Fed staff forecasts, but only an input to the overall forecasting process at the Bank.

But the bottom line is that while human forecasts, while always wrong, have an interval of certainty around the predicted, if very wrong, number the Fed's DSGE model is far more nuanced.

Uncertainty around the forecasts is significant, particularly for GDP growth. The width of the 68 percent probability interval for GDP growth is 3.8 percentage points in 2015 and widens to 5.3 percentage points in 2017. The 68 percent probability intervals for inflation remain relatively tight, ranging from 0.4 percent to 1.3 percent in 2015 and from 0.4 percent to 2.1 percent in 2017.

What this means is that while carbon-based economists can and will predict the future and only after the fact blame the weather (which was accessible to everyone and quite public at the time the predictions were made), for being massively wrong, robots take the other approach: they have a ridiculous range which at least assures that they will be correct. Then again, as the chart below shows, while the Fed's DSGE model forecasts GDP of either -4% or +8% in 2017, that will hardly provide comfort to the millions of Americans caught in the latest depression if, as usually happens, the bearish prediction turns out to be accurate.

And this is what the DSGE model "predictions" look like when visualized:

Source: The FRBNY DSGE Model Forecast--April 2015