“based on historical relationships, the recent declines in equity valuations indicate that we are currently in a situation of somewhat heightened risk of a recession – but at this time it is in a gray area that does yet outright indicate a recession compared to what has been observed historically”.
I provided an analysis of the current state of the business-cycle in my previous entry (briefly: the industry sector is subject to contraction, but there is only weak evidence that the sector-dynamics have spread over the whole economy, as of now). Temptingly, in this entry, I’d like to revert the (Granger-) causality implied by John’s statement: instead of inferring recession-probabilities from financial data, I’d like to suggest asset-management opportunities obtained from (macro-) economic indicators.
- After reader-feedback I here provide the daily indicator series (designs 1-4) for further analysis: indicator_system.txt (the series start in April 1997).
If today marked the onset of a recession, then this event would be at odds with the historical track with respect to
- oil prices (low)
- interest rates (low)
- job creation (high)
A constellation unseen before. But each recession is different. So: what does the data tell us, right now, about the eventual onset of an `at odds’ recession? Continue reading
In yesterday’s entry 1 I posted a new daily indicator. I deliberately overemphasized the business-cycle in its design. What does this mean, you may ask? The following graph compares yesterday’s indicator, blue line, with a (new daily) `neutral’ indicator, red line, which tracks `growth’ of the industrial production series (the business-cycle is not overemphasized):
Growth as measured by the `neutral’ indicator is not negative, yet. Continue reading
I noticed, recently, worries about the possibility of an up-coming recession in the US, see 1 (and the accompanying puzzle 2). It’s been a long time since my last entry on the topic… Sufficient time, indeed, for the business-cycle to eventually complete. And sufficient time for me to work on a new mixed-frequency extension of the MDFA, 3. As fruit of these efforts let me here share today’s up-date of the new daily real-time indicator. Continue reading
In my previous entry I proposed a new MDFA-extension in the case of mixed-frequency data, see this presentation. In the following, I’d like to analyse the estimation problem and to contrast my findings with classic time-domain approaches, as discussed in Foroni and Marcellino (concise and comprehensive review of the sota in that matter). To conclude, I briefly discuss the MDFA-solution presented in my slides. Continue reading
I give a talk on a brand-new MDFA-feature at the up-coming CFE-CMStatistics in London. Specifically, I `show’ (glance/glimpse) how to mix/combine data sampled at different frequencies in the generic MDFA-framework. As a result, I derive two daily indicators for US-GDP and for US-IPI, based on monthly (INDPRO, UNRATE, PAYEMS), weekly (ICSA) and daily (S&P500, VIX) data. A simple trading strategy (long if the resulting macro-indicators are positive, short if negative) applied to the S&P500 illustrates the timing-properties of both real-time indicators over the last two recessions. Here’s my presentation (click for download).
I received multiple feedbacks about line overflows in MDFA-Legacy. I’ll check this issue next month (it’s mostly a LaTex problem or, to be more specific, a problem I have with LaTex). I will poste the newest version of my MDFA-code too, which tackles regularization (I disclosed regularization in text-form, see chapter 10 in MDFA-Legacy, but I did not poste code yet: so that’s up-coming too).
Since (slightly more than) one year my `spare’ time (non-teaching duties) is dedicated to trading applications of MDFA. These activities were/are time-absorbing; sufficiently so to hinder any up-date of my running MDFA-Legacy project. However, I feel confident that I’ll be able to resume activity in the context of that project next year. I intend to emphasize the chapters on the various forms of non-stationarity, including unit-roots, cointegration, adaptive filters and data revisions (the latter being interpreted as a particular non-stationarity along the data-vintages).