Timing of the Money Effect - Implications for Monetary Policy

In order to assess, understand and optimize MP, one has to derive the parameters and ki-netics of its MTC effects in monetary transmission processes and mechanism. This is essen-tially important to feed new or the existing predictive MP models with the correctly deter-mined variables, and time or timing plays of course a pivot role here. As anticipated in di-verse MP models (see 3.2.3) timing refers to the fact that a MP decision is made at any time along the time scale and each decision point in time has overlapping effect along the time line with all of the effects of the other decision point - leading to a huge overlap of causes and effects over time - that transmit into the real economy via MTCs (see 4.3.8).

Page | 126

Among all model approaches to estimate the impact and timed influence of a change in money supply (Δ%M0) on the real economy a correlation model of price stability (Δ%HICP) and economic growth (Δ%GDP) was performed to roughly model the key dynamical trends. Today there is only little knowledge about the precise timing of these complex effects. To solve this complex problem with a new approach - as the VAR analysis models (see 3.2.3) that are widely used by monetary experts and scientists have some draw-backs - an alter-native approach was chosen here: ‘a Time-Matrix Pearson Correlation Time Lag Test’. This tests, seemingly previously also already known as ‘dynamic correlation studies (DC)’ (Walsh 2010 and references herein) was performed with the new consolidated ECB data to carve out the respective time windows of effects to a series of MP change-steps. The weighted overlay of all serial effects yields a prediction or description model of MTC effects, here focusing on prices and GDP growth due to a requirement to simplify. Figure 49 provides the result of a 16-year DC study conducted on %ΔM3 on %Δ HICP, real and nominal %ΔGDP.

Figure 49 DC Test for EMU M₃, Inflation, real and nominal GDP (1999-2015)

For real GDP (red line), in a 15 year projection, a ‘π/4-cosine sinusoidal reminiscent har-monic oscillation effect of R for M3 on economic growth becomes apparent and is more distinct than the effect of M0, because M3 is further downstream of the monetary transmis-sion channel and represents the real economy’s ‘available liquid assets’ for investment and consumption. To some extent this also hold true for nominal GDP (blue line) but the net effect is reduced due to the price channel effects. The timed effect of HICP shows also a sinusoidal reminiscent harmonic oscillation but at a 1.5-fold lower frequency (green line).

Page | 127

Put simply, a punctual monetary stimulus has on average a slight positive impact on GDP and inflation in the very short run, followed by a slight negative in year one and two. After three years it has again a positive effect and the effects decompose later on due to a differ-ent frequency see Figure 49. The strongest correlation on prices was seen after four to five years. Signal-to-noise-tests (t-test) reveal that a higher DC correlation also indicates a prob-able stronger effect: a forecast or impulse response function (IRF) is the sum of all effects.

An increase of M3 has a direct and immediate effect on GDP, real GDP and nominal GDP, while prices are more rigid until year three after a quantity stimulus (the media MP channel is not included here, only the quantity changes of M3 - other effects could be modeled in).

The strength of the predictability geometrically declines with the forecasted number of future years and thus also the error based on the correlation and probable strength of the effect. Nevertheless, the model clearly predicts that an inflationary period is followed by a deflationary phase at a high confidence interval (p<0.05), e.g. at seven to eight years post-M-stimulus. Simply due to the oscillatory behavior the overall averaged effect cancels out to almost 0: the classical notion that money is neutral in the long-run. However, the effect declines with time geometrically, meaning if MP finds the right ‘time windows’ to stimulate the economy with new money and then - and only then - it can turn positive and drive GDP.

Compared to a VAR analysis prediction it is likely to be more robust to shocks. VAR analysis are more affected by shocks like the FC+EC, and other runaway values. As a result well-adjusted p-VAR more pessimistically predicts GDP growth in the long-run at steady prices (6 Figure S3), due to the periodic occurrence of shocks [real GDP and price VAR prediction both turn relatively optimistic without the crises, which leads to ambiguous interpreta-tions]. Although it yields too pessimistic prospect for future real GDP it still relinquishes a periodic and oscillatory ‘sinusoidal nature’ in the future time-line. This was also divulged as a common feature of the DC method, further corroborating that MP decisions (per time intervals) cause sinusoidal aggregate MTC waves of (1) different frequency and (2) ampli-tude, and (3) depending on output a different shift of the oscillatory behavior.

Thus, MP stimulates the amplitudes of MTC waves and timing is the key to its net benefits.

The Dynamic Time Matrix Pearson Correlation Forecast is probably new: it is less prone to errors and runaway values and based on multiple VAR-chain extensions of R values. It is also more flexible and adaptive. A DC-forecast for the EMU is given in (6 Figure S4). If the

Page | 128

growth of money were knowable, it could provide an even more powerful forecasting tool.

It can be also used to extract diverse DC-forecast-coefficient matrixes of MP effects on out-put variables - also by comparing historic periods: for instance the effect of M₃ on HICP in the Pre-Euro time (1991-2003) has had a significantly different R-matrix (see 6 Figure S5).

With regard to recent ECB €1tnQE program (60bn/m) the new DC results advise a more careful raising of EMU base money due to the four to five year delayed effect on prices.