There was overwhelming relief that the September jobs report, which came out this week, showed a number in the mid-200k's. That's it? Why should there be relief, if everyone genuinely believed that the "bad" August number was simply an outlier, and that the high-200k's jobs gain streak that we had (for a semi-year prior to that) was simply intact? We noted in our web log article a couple days ago (incidentally a top news read on Bloomberg View), that the story here is still more complicated and not simply one of an "outlier" in an otherwise optimal trend. In this note we will provide just quick additional reflections on this topic.
I have been a proponent of focusing on and increasing job growth (one can also see throughout my web log that there are many articles examining relationships within employment and unemployment ratios). Anyone desiring work should be able to find something, at least partially, rewarding. In the U.S., we analyze monthly jobs data by the thousands, so small numbers with these (000s) units unfortunately just becomes unidentifiable statistics. It's even worse when we dicker about guesses on the basis point of the unemployment rate. But we know that having significant net increases in job opportunities is healthy. It not only enhances people's livelihoods and retainable skill set across all socioeconomic groups (not just within isolated pockets of wealthy shareholders), but the steady paycheck also provides a meaningful platform for esteem and self-worth.
Still, we need to be true to the real changes happening in economic trends, regardless of where they head. We should all appreciate that these labor numbers don't just hover at a fixed 250k or so, particularly throughout the late stages of a growth cycle. As anyone can appreciate, similar to the last couple of times the unemployment rate has crossed 5.9% (either forecibly on the way up or as we see wiggling on the way down), today it is still a big struggle for many job seekers to find reasonable employment. Making ends meet for those in the bottom 2/3 continues to be more challenging since before the recent recession (one can read the earlier link).
In the chart below, we recreate the top chart from our prior web log article (we updated though the incoming revisions noted on the September labor report). With grey dashed lines, we then superimpose a simple kernalized, two-month average. This simple two-step binomial model is less biased than the traditional, backward-facing moving average. Instead our approach will triangulate labor information +1 month, capturing the basic essence of where the data is at any given month. We've been very clear with our math and still lose no overall information either, as both the beige vertical bars and the grey dashed line, average the same 218k in the chart below.
We also showed in our earlier article that the typical standard deviation (σ) is ~100k. For our kernalized average, which is essentially a width of a 2-month average, it is a little less at ~70k:
The value of this σ shows that there is a fairly normal, non-outlier slipping pattern in the U.S. economy. We should certainly desire for something more, particularly for the many millions who should be working. But from a statistics perspective we couldn't expect it. Some pundits may now again try to force too simple math on this labor data, especially as we drive quickly toward mid-term elections. The problem with such oversimplification is that is trivializes the subtle nuances that are pertinent. We've seen over the past month how economists and media pundits have tried to force through with such ease, explanations for the labor path that has completely failed to explain recent underlying shock patterns that we have seen. And not just in August 2014, but also in December 2013. The basic concept of glossing over fundamental details was central to these famous comments from more than 80 years ago, at an Oxford lecture:
It can scarcely be denied that the supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience.
Words delivered through a knowing theoretical physicist, Albert Einstein.
I have been a proponent of focusing on and increasing job growth (one can also see throughout my web log that there are many articles examining relationships within employment and unemployment ratios). Anyone desiring work should be able to find something, at least partially, rewarding. In the U.S., we analyze monthly jobs data by the thousands, so small numbers with these (000s) units unfortunately just becomes unidentifiable statistics. It's even worse when we dicker about guesses on the basis point of the unemployment rate. But we know that having significant net increases in job opportunities is healthy. It not only enhances people's livelihoods and retainable skill set across all socioeconomic groups (not just within isolated pockets of wealthy shareholders), but the steady paycheck also provides a meaningful platform for esteem and self-worth.
Still, we need to be true to the real changes happening in economic trends, regardless of where they head. We should all appreciate that these labor numbers don't just hover at a fixed 250k or so, particularly throughout the late stages of a growth cycle. As anyone can appreciate, similar to the last couple of times the unemployment rate has crossed 5.9% (either forecibly on the way up or as we see wiggling on the way down), today it is still a big struggle for many job seekers to find reasonable employment. Making ends meet for those in the bottom 2/3 continues to be more challenging since before the recent recession (one can read the earlier link).
In the chart below, we recreate the top chart from our prior web log article (we updated though the incoming revisions noted on the September labor report). With grey dashed lines, we then superimpose a simple kernalized, two-month average. This simple two-step binomial model is less biased than the traditional, backward-facing moving average. Instead our approach will triangulate labor information +1 month, capturing the basic essence of where the data is at any given month. We've been very clear with our math and still lose no overall information either, as both the beige vertical bars and the grey dashed line, average the same 218k in the chart below.
Does the basic pattern appear steady or rising? Does it even look as if in September we have properly rebounded to anything similar to the glorious six "others" months from earlier this year? The answer to both should be no. In fact, the average we have for September is still below that from July. And we'll see later it is well below that from the other earlier months. The Wall Street Journal stated in a recent print article that "as even a child could see, though, that (August) figure didn't look like the others." Today an MIT professor friend of mine, Seth Hannah, had phoned me about this not-so-correct article. Let's look again at the illustration above and try this again: Would a child really describe the multi-month trend, since the end of Q1, as mostly steady-to-up?
Let's focus a little deeper into this rhetorical question. Staying now with just averages, the current average of 225k is nearly 40k-50k lower than it was earlier this year. Since the end of Q1, we've been steadily sliding lower, for much of the past 5 months. Sure 225k is a boost from the sub-200k we saw in August, but this is the nature of what we already mentioned in our prior web log article. Temporary shocks, in both directions, often occur during some economic slowing. We can also see in the scatter plot, the second illustration from our prior article, that if one uses a prior month (August) read along the horizontal-axis of high-100k's, we would still get an expected following month (September) vertical-axis read in the mid-200k's. So this week's September read of 248k is hardly that large of a positive rise versus the otherwise recent slip in labor data.
We also showed in our earlier article that the typical standard deviation (σ) is ~100k. For our kernalized average, which is essentially a width of a 2-month average, it is a little less at ~70k:
σ/√n
Where n=2, for number of months.
Where n=2, for number of months.
The value of this σ shows that there is a fairly normal, non-outlier slipping pattern in the U.S. economy. We should certainly desire for something more, particularly for the many millions who should be working. But from a statistics perspective we couldn't expect it. Some pundits may now again try to force too simple math on this labor data, especially as we drive quickly toward mid-term elections. The problem with such oversimplification is that is trivializes the subtle nuances that are pertinent. We've seen over the past month how economists and media pundits have tried to force through with such ease, explanations for the labor path that has completely failed to explain recent underlying shock patterns that we have seen. And not just in August 2014, but also in December 2013. The basic concept of glossing over fundamental details was central to these famous comments from more than 80 years ago, at an Oxford lecture:
It can scarcely be denied that the supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience.
Words delivered through a knowing theoretical physicist, Albert Einstein.
No comments:
Post a Comment