- by New Deal democrat
As readers know, I am trying to reverse engineer the ECRI Weekly Leading Index.
Research has revealed that the WLI (per Lakshman Achuthan) has 7 components. We know that two are initial jobless claims and the JoC-ECRI commodity index. Six of the seven are public record.
Since the possible early monthly reports were out last week, they wouldn't count in any event. Unfortunately, I don't have time to be as thorough as I would normally like, but ALL of the 6 public weekly components (with the possible exception of the rate of growth of real estate loans, I haven't had time to check) were positive this week:
DJ Bond Avg up
S&P 500 up
Initial Jobless claims*: down over 40,000
Commodity price changes: up
Purchase Mortgage applications: up
10 year treasury - BAA credit spread*: down
The predicted week over week change is therefore predicted to be positive.
[*Note: these are inverse relationships, so the higher the number, the lower the growth score]
We'll have an answer tomorrow.
Showing posts with label Weekly Leading Index. Show all posts
Showing posts with label Weekly Leading Index. Show all posts
Thursday, January 19, 2012
Monday, January 16, 2012
The ECRI Weekly Leading Index, unmasked
- by New Deal democrat
This is a continuation of my series reverse engineeering ECRI's Weekly Leading Index (WLI). I began by noting that ECRI's founder, Prof. Geoffrey Moore, in 1990 initially proposed a Weekly Leading Index made up of indicators whose values could be computed weekly, or if monthly were reported at the beginning of the next month. While this index would be a little less reliable than the Long Leading and Short Leading Indeces, it would have the advantage of being more timely.
Its weekly components were:
Real M2
Dow Jones Bond Average
S&P 500 stock price index
Initial claims for unemployment insurance
Journal of Commerce change in commodity prices
Dun and Bradstreet new business formation and large business failure
Real estate loans, deflated, growth rate
The quickly reported monthly components were:
Average workweek in manufacturing
Layoff rate under 5 weeks
ISM manufacturing vendor performance
ISM manufacturing inventory change
The questions remained, were the monthly components included in the final result? Were the weekly components weighted? Were there any changes in the list? As to the last, I have noted that it is almost impossible to generate a decline such as ECRI claims for the WLI in 2011 if Real M2, which experienced a tsunami-like increase in late summer, were still included. I have suggested that the credit spread between government and corporate bonds might have replaced Real M2 in the list at some point.
It appears that ECRI spokesman Lakshman Achuthan himself has all but settled the issue. Appearing as a guest author at Barry Ritholtz' "The Big Picture" blog in 2009, he wrote:
The link takes you to the book "Beating the Business Cycle," authored by Achuthan in 2004. There, at pp. 108-09 he says
So that is pretty clear. There are seven indicators (confirming that the early monthly reports are not part of the index), and specifying that initial jobless claims and the JoC-ECRI industrial commodity index are two of the components.
Finally, a little google searching reveals that several issues of Business Week including the "Business Week Leading Index" are archived online. Here is a screenshot of one of them, edited only to delete a line describing another unrelated index. Note that it includes precisely seven indicators:
Another example can be found here.
So that just about settles it. The referenced "real estate loans" are sourced to the Federal Reserve Bank and weekly updates can be found here.
It appears that the index is weighted. Also, I continue to question whether Real M2 survived the transition, just as it appears that AAA bonds were substituted for the similar Dow Jones Bond Average. With those caveats, the components of the WLI are unmasked.
This is a continuation of my series reverse engineeering ECRI's Weekly Leading Index (WLI). I began by noting that ECRI's founder, Prof. Geoffrey Moore, in 1990 initially proposed a Weekly Leading Index made up of indicators whose values could be computed weekly, or if monthly were reported at the beginning of the next month. While this index would be a little less reliable than the Long Leading and Short Leading Indeces, it would have the advantage of being more timely.
Its weekly components were:
Real M2
Dow Jones Bond Average
S&P 500 stock price index
Initial claims for unemployment insurance
Journal of Commerce change in commodity prices
Dun and Bradstreet new business formation and large business failure
Real estate loans, deflated, growth rate
The quickly reported monthly components were:
Average workweek in manufacturing
Layoff rate under 5 weeks
ISM manufacturing vendor performance
ISM manufacturing inventory change
The questions remained, were the monthly components included in the final result? Were the weekly components weighted? Were there any changes in the list? As to the last, I have noted that it is almost impossible to generate a decline such as ECRI claims for the WLI in 2011 if Real M2, which experienced a tsunami-like increase in late summer, were still included. I have suggested that the credit spread between government and corporate bonds might have replaced Real M2 in the list at some point.
It appears that ECRI spokesman Lakshman Achuthan himself has all but settled the issue. Appearing as a guest author at Barry Ritholtz' "The Big Picture" blog in 2009, he wrote:
Our leading indexes are composites of key drivers of the business cycle. Correlations are not part of our process which focuses on the relationship of indicators around inflection points in growth and inflation. .... ECRI does not suggest that the LLI is a perfect leading indicator, but it does not include stock prices and has a longer lead than stock prices over growth rate cycle turns.(my emphasis) (UPDATE: To avoid confusion, please note that the reference to stocks not being included, is to ECRI's "long leading index," a completely separate index).
You can read more about the Weekly Leading Index (formerly known as the Business Week leading index) here: http://books.google.com/books?id=vSz99DDF-q8C&pg=PA107&dq=beating+the+business+cycle+weekly+leading+index&client=firefox-a
The link takes you to the book "Beating the Business Cycle," authored by Achuthan in 2004. There, at pp. 108-09 he says
"To monitor developments on a more frequent basis, Moore helped develop a Weekly Leading Index (WLI) in 1983. Some of the components of the WLI, like initial jobless claims, had long been available on a weekly basis. Others had to be created from scratch, like the Journal of Commerce-ECRI industrial materials price index, designed to measure inflation in a broad range of industrial raw materials. Starting in 1983, the weekly index was published in Business Week magazine. Known at the time as the Business Week Leading Index, ... In the late 1990s, ECRI began to publish the index through our own website, businesscycle.com.Then, at pp. 135-36 he elaborates:
"We .. determin[e] good proxy measures for the various drivers [of the economy] -- that is, individual leading indicators -- and combin[e] them into composite indeces which objectively summarize their information. For example, in constructing the Weekly Leading Index (WLI) , we use a specific leading indicator -- initial claims for unemployment insurance -- to represent employment, which is a key driver of the business cycle. Likewise, we pick out six other specific leading indicators that are updated weekly, to represent other cyclical forces. Because the seven indicators are rarely unanimous, we summarize them into the composite WLI ...."
(my emphasis)
Finally, a little google searching reveals that several issues of Business Week including the "Business Week Leading Index" are archived online. Here is a screenshot of one of them, edited only to delete a line describing another unrelated index. Note that it includes precisely seven indicators:
Another example can be found here.
So that just about settles it. The referenced "real estate loans" are sourced to the Federal Reserve Bank and weekly updates can be found here.
It appears that the index is weighted. Also, I continue to question whether Real M2 survived the transition, just as it appears that AAA bonds were substituted for the similar Dow Jones Bond Average. With those caveats, the components of the WLI are unmasked.
Thursday, January 12, 2012
Beta testing the Shadow Weekly Leading Index
- by New Deal democrat
As readers know, I am trying to reverse engineer the ECRI Weekly Leading Index. Prof. Geoffrey Moore recommended 11 elements for this index when he proposed it in 1990. As originally conceived, the weekly index was supposed to anticipate the monthly LEI, which had to rely on late reports like housing permits and durable goods orders. By contrast, the 4 monthly numbers incorporated into the weekly index would all be known within one week of the end of the previous month. The weekly index would be slightly less reliable than the LEI, but by dint of early report, would be very useful.
Ten of the 11 elements are publicly reported (the 11th is Dun and Bradstreet's weekly number of business formations and dissolutions which is not available to the public). Of the remaining 10, six are weekly series and the other 4 are the monthly series noted above. I have already created several preliminary graphs based on only 7 elements which look very close to ECRI's Weekly Index, including its roller coaster ride of the last two years. So it's time to beta-test and make refinements based on the results.
The list below is the changes in those 10 numbers for the last week. The only change I have made is the substitution of credit spreads for real M2, which I have reason to believe was replaced on the list. Another issue is whether ECRI continues to include the early monthly reports, or changed Moore's original concept by only relying on weekly reports. The last issue is whether the report is an unweighted average of the elements or not. I am giving two forecasts for the WLI, one that includes and one that does not include the early monthly reports. The forecast assumes an unweighted index.
The following weekly components changed as follows for the week ending January 6 (YoY change in parenthesis):
DJ Bond Avg -.07 to 114.58, or -0.1% (YoY up from 111.12, or +3.1%)
S&P 500 +20.21 to 1277.81, or +1.6% (YoY up from 1271.50, or +0.5%)
Initial Jobless claims*: +27,000 to 399,000 or -6.8% w/w (YoY -15,000 from 414,000, or +4.8%)
Commodity price changes: +2.47 to 119.81, so +2.1% (YoY -17.05 from 133.65, or -10.4%)
Purchase Mortgage applications: +8.1% w/w (YoY -17.9%)
10 year treasury - BAA credit spread*: +.02 to 3.29%, or -0.6% (YoY up from 2.69%, or -18.2%)
If only weekly series are used, the predicted week over week change is +0.7, and the YoY growth rate is -6.4.
The following early monthly reports changed as follows (YoY change in parenthesis):
Avg manufacturing workweek +.1 to 41.5, or +0.2% (+.2 YoY, or +0.5%)
Unemployment 0-5 weeks*:+159 to 2669, or -6.0% (YoY -32, or +1.2%)
ISM vendor performance flat at 49.9, or 0.0 (YoY -6.8, or -12%)
ISM inventory change -1.2 to 47.1, or -2.5% (YoY -4.7 or -9.1%)
If the early monthly reports are included, the unweighted predicted weekly change is -0.4 and the YoY growth rate is -5.8.
[*Note: these are inverse relationships, so the higher the number, the lower the growth score]
We'll have an answer tomorrow. I anticipate refining the forecast as I am able to determine better how ECRI constructs their Index.
As readers know, I am trying to reverse engineer the ECRI Weekly Leading Index. Prof. Geoffrey Moore recommended 11 elements for this index when he proposed it in 1990. As originally conceived, the weekly index was supposed to anticipate the monthly LEI, which had to rely on late reports like housing permits and durable goods orders. By contrast, the 4 monthly numbers incorporated into the weekly index would all be known within one week of the end of the previous month. The weekly index would be slightly less reliable than the LEI, but by dint of early report, would be very useful.
Ten of the 11 elements are publicly reported (the 11th is Dun and Bradstreet's weekly number of business formations and dissolutions which is not available to the public). Of the remaining 10, six are weekly series and the other 4 are the monthly series noted above. I have already created several preliminary graphs based on only 7 elements which look very close to ECRI's Weekly Index, including its roller coaster ride of the last two years. So it's time to beta-test and make refinements based on the results.
The list below is the changes in those 10 numbers for the last week. The only change I have made is the substitution of credit spreads for real M2, which I have reason to believe was replaced on the list. Another issue is whether ECRI continues to include the early monthly reports, or changed Moore's original concept by only relying on weekly reports. The last issue is whether the report is an unweighted average of the elements or not. I am giving two forecasts for the WLI, one that includes and one that does not include the early monthly reports. The forecast assumes an unweighted index.
The following weekly components changed as follows for the week ending January 6 (YoY change in parenthesis):
DJ Bond Avg -.07 to 114.58, or -0.1% (YoY up from 111.12, or +3.1%)
S&P 500 +20.21 to 1277.81, or +1.6% (YoY up from 1271.50, or +0.5%)
Initial Jobless claims*: +27,000 to 399,000 or -6.8% w/w (YoY -15,000 from 414,000, or +4.8%)
Commodity price changes: +2.47 to 119.81, so +2.1% (YoY -17.05 from 133.65, or -10.4%)
Purchase Mortgage applications: +8.1% w/w (YoY -17.9%)
10 year treasury - BAA credit spread*: +.02 to 3.29%, or -0.6% (YoY up from 2.69%, or -18.2%)
If only weekly series are used, the predicted week over week change is +0.7, and the YoY growth rate is -6.4.
The following early monthly reports changed as follows (YoY change in parenthesis):
Avg manufacturing workweek +.1 to 41.5, or +0.2% (+.2 YoY, or +0.5%)
Unemployment 0-5 weeks*:+159 to 2669, or -6.0% (YoY -32, or +1.2%)
ISM vendor performance flat at 49.9, or 0.0 (YoY -6.8, or -12%)
ISM inventory change -1.2 to 47.1, or -2.5% (YoY -4.7 or -9.1%)
If the early monthly reports are included, the unweighted predicted weekly change is -0.4 and the YoY growth rate is -5.8.
[*Note: these are inverse relationships, so the higher the number, the lower the growth score]
We'll have an answer tomorrow. I anticipate refining the forecast as I am able to determine better how ECRI constructs their Index.
Subscribe to:
Comments (Atom)