Real M1 growth in the euro area has been moderating in recent quarters, adding to concerns about the economic outlook given the robust relationship between the business cycle and narrow money. This box shows that the leading and pro-cyclical properties of real M1 for real GDP remain a robust stylised fact in the euro area. Moreover, there are indications that these properties reflect the predictive capacity of narrow money, beyond the influence of interest rates. At the current juncture, models exploiting the predictive power of real M1 suggest that the steady decline in real M1 growth from its most recent peak in mid-2017 points to very low risks of recession in the euro area up to the beginning of 2020.
The leading and pro-cyclical properties of real M1 with respect to real GDP in the euro area remain a robust stylised fact. These properties, which can be found for the relationship between real narrow money and real economic activity in both levels and growth rates, have been documented in various publications for earlier time periods. An illustration of such properties can be derived from a visual examination of monthly data from January 1970 to February 2019 for annual growth in real M1, which is defined as the nominal narrow money aggregate M1 deflated by the HICP. Specifically, it is notable that this growth rate went well into negative territory for prolonged periods just before (or in coincidence with) all historical euro area recessions, as dated by the CEPR Euro Area Business Cycle Dating Committee (see Chart A).
Real M1 annual growth and euro area recessions
(annual percentage changes)
Sources: CEPR, ECB.
In terms of turning points in the levels of real M1 and real GDP, statistical indicators suggest that the lead and pro-cyclicality of peaks and troughs in real M1 in relation to peaks and troughs in real GDP represent a historical regularity. Indeed, concordance indices calculated at different leads and lags indicate that turning points in real M1 tend to lead turning points in real GDP by four quarters, on average, and that, with that lead for narrow money, real M1 and real GDP are estimated to spend almost 90% of the time in the same business cycle phase (see Chart B). Moreover, the strong degree of synchronisation between turning points appears to have remained stable since the 1970s, including over more recent sub-periods.
Concordance indices between real GDP and real M1 at different leads and lags
Sources: CEPR, ECB.
Empirical evidence suggests that the predictive power of real M1 for real output in the euro area is not simply a reflection of information contained in the yield curve. Against the background of the ample evidence pointing to the leading properties of the slope of the yield curve for predicting recessions, it is natural to ask to what extent the leading and pro-cyclical properties of real M1 are driven by the pattern of the yield curve – and to what extent controlling for yield curve constellations would obviate narrow money as a predictor for economic activity. Historical data for the euro area suggests that the slope of the yield curve exhibits a positive co-movement with respect to real M1, with a lead of two quarters. In line with this historical regularity, concordance indices at different leads and lags indicate that turning points in the slope of the yield curve tend to lead turning points in real GDP by six quarters, on average, and that, with that lead, the slope of the yield curve and real GDP are estimated to spend about 60% of the time in the same business cycle phase (see Chart C). This regularity reflects the fact that interest rates do indeed represent key variables in driving portfolio decisions by households and non-financial corporations. However, it does not necessarily imply that the predictive power of real M1 for real economic activity in general, and for recession risks in particular, is entirely driven by its relation to the yield curve. In fact, the concordance indices with respect to real GDP point to a stronger degree of co-movement with real M1 (with a one-year lead, as depicted in Chart B) than with the slope of the yield curve (with a one-and-a-half-year lead, as depicted in Chart C). Moreover, the relationship between real M1 and real GDP as measured by these indices appears more stable over sub-periods than that between the slope of the yield curve and real GDP.
经验证据表明，实际M1对欧元区实际产出的预测能力不仅仅是收益率曲线中包含的信息的反映。在充分证据指出收益率曲线斜率的领先性质预测衰退的背景下，很自然地会问到实际M1的领先和顺周期特性在何种程度上受到收益率曲线模式的驱动。 - 控制收益率曲线将在多大程度上避免将狭义货币作为经济活动的预测因素。欧元区的历史数据表明，收益率曲线的斜率与实际M1呈现正向共同运动，领先两个季度。根据这种历史规律，不同领先和滞后的一致性指数表明，收益率曲线斜率的转折点往往领先实际GDP的转折点平均达到六个季度，并且在此基础上，估计收益率曲线和实际GDP在同一商业周期阶段约占60％的时间（见图C）。这种规律性反映了这样一个事实，即利率确实是家庭和非金融公司推动投资组合决策的关键变量。然而，它并不一定意味着实际M1对一般实际经济活动的预测能力，尤其是衰退风险，完全取决于其与收益率曲线的关系。事实上，与实际国内生产总值相关的一致性指数表明，与实际的M1（具有一年的领先优势，如图B所示）的联动程度要强于与收益率曲线的斜率（具有一年半的领先优势 - 如图C所示）。此外，通过这些指数衡量的实际M1与实际GDP之间的关系在子期间看比在收益率曲线的斜率与实际GDP之间更稳定。
Concordance indices between real GDP and the slope of the yield curve at different leads and lags
Sources: CEPR, ECB.
Turning to the current juncture, a formal econometric analysis based on probit models exploiting the predictive power of real M1 does not point to significant recessionary risks in the euro area for 2019 and early 2020. On the basis of data since 1970, the probability of a contraction in euro area real GDP derived from a probit model based on real M1 (lagged by 12 months) increased sharply before all previous euro area recessions (see Chart D), providing strong evidence of the usefulness of narrow money in predicting recessions in the euro area. Forecasts based on this model point to recession risks increasing slightly in 2019, from about 1% in January 2019 to between 5% and 7% in the second half of 2019 before falling to below 5% in February 2020, that is to say remaining very low (blue line). Controlling for the slope of the yield curve changes results only marginally (yellow line). Overall, the current level of real M1 growth is still comfortably above the zone that would be associated with risks of a recession in the near future.
Euro area recession probabilities based on probit models with lagged real M1
Sources: CEPR, ECB.