Last week, using nineteen economic indicators, I showed the current state of the economy on a quadrant grid. This week, I’m continuing the analysis by comparing today’s economy with conditions prior to the 2007 and 2001 recessions. Investors should find the then-and-now comparisons informative.
At the end of the article, after the grid comparisons, I have some explanations about the process and assumptions.
Importantly, for anyone who is familiar with the graphics in my last article, I have switched the axes on the grid. The baseline is now the vertical axis, which places the baseline line horizontally across the grid. This makes more sense visually as the indicators are plotted either above or below the baseline. In addition, I have adjusted the annualized rate of change for all indicators so that when they are mapped on the grid the rate of change is proportional (see the Adjusted Rate of Change section towards the end of this article for a more detailed explanation).
The economic indicators used in the analysis are ones that historically turn down before or at the start of a recession. For some measures that are considered coincident indicators, like retail sales, when adjusted for inflation and/or population, they act more like leading indicators.
Below are descriptions of the grid quadrants (again, these have changed from my last article because I switched the axes).
Right, upper quad: Above baseline, with positive trend
Left, upper quad: Above baseline, with negative trend
Right, lower quad: Below baseline, with positive trend
Left, lower quad, Below baseline, with negative trend
There is always some lag in gathering economic data. The data used here to show the “current economy” is from August, or, in the case of job openings and hires, the end of July.
Table 1 and 1a show the nineteen indicators that are mapped on the Recession Indicators – Current Economy grid. The Mean of Coordinates ((MoC)) is the averages of two axes coordinates. When plotted, the MoC turns out to be the most useful point on the grid. It serves as an indication of the economy’s overall health at a point in time, which is especially useful when comparing grids.
Because the yield curve spread and STLFSI data include negative numbers and they have a large range of change, it is difficult to calculate comparative percentages for the grid. Consequently, for these two measures, I’m approximating their locations. Neither of these two indicators is used to calculate the MoC.
As can be seen in Grid 1, seventeen of the nineteen indicators are above the baseline, and the majority of the seventeen show a positive annualized rate of change. Consequently, the MoC (red square) is positioned favorably: plus 14% above the baseline and plus 7% rate of change.
Economy One Month Prior to 2007 Recession
Table 2 and 2a show the ones that are mapped on the Recession Indicators – 1 Month Prior 2007 Recession grid (Grid 2). The indicators are the same as Table 1, except for the ISM non-manufacturing composite index. Although ISM began reporting on non-manufacturing data in 1998, the non-manufacturing composite index began January 2008.
In Grid 2, one month prior to the official start of the 2007 recession, most of the indicators are in the negative rate-of-change quadrants, both above and below the baseline. The MoC is on the negative rate of change side and nearly on the baseline, which is what should be expected. This suggests that the baseline values I’ve selected must be reasonably accurate as a whole.
Although industrial capacity utilization, industrial production, and real retail sale per capita are above the baseline with a positive rate of change, they are marginal, at best. Yet, this is one month before the recession officially started. As the recession continued, they all nosedived.
What is interesting to see is the distance the indicators would have to travel from where they are in Grid 1 to get to where they would be at the start of a recession, as demonstrated in Grid 2. The current economy doesn’t appear to be anywhere near a recession.
I should also point out the yield curve spread is a contrary indicator at this point. The spread had inverted in August 2006, nearly fifteen months earlier. The spread became positive again in June of 2007 and began to steadily increase. The yield curve spread has typically inverted 12 to 15 months prior to recessions, and then becomes positive before the start of a recession. Where we see the yield curve spread on Grid 2 is no surprise.
Economy Six Months Prior to a 2007 Recession
I thought it would also be interesting to see what the economy looked like six months prior to the 2007 recession. The same indicators are used here as in Table 2.
In Grid 3, at six months prior to a recession, there has been a clear shift to the “dark side.” None of the indicators are more than 21% above the baseline, a marked difference from the current economy grid (Grid 1). In addition, ten of the eighteen indicators are showing negative rates of change. This results in an MoC that has a negative rate of change and is moving towards the baseline. I find it interesting that the MoC is in the negative rate of change grid at this point. I need to do more analysis to find the time frame (nine to 12 months?) at which the MoC crosses over into the negative quad. The MoC crossing over to the negative rate of change quad may be the point at which a recession is inevitable.
This graphic is useful because it shows that even six months prior to a recession, the grid mapping gives a firm warning of a pending downturn.
The Economy 1 Month Prior to the 2001 Recession
By mapping the beginning of the 2001 recession, a comparison can be made to conditions before the 2007 recession. Unfortunately, as shown in Table 4, I don’t have as much data for that time period. However, there still are a number of sound indicators available and the mapping turns out to be useful.
There are some similarities between Grid 4 and Grid 2, but also some interesting differences. Similarities include the positioning of most indicators in the negative rate of change quadrants and a number of indicators being below the baseline. In addition, the MoC is near the baseline and shows a significant negative rate of change. One difference is that the indicators in this grid are more spread out than in Grid 2. Another difference is that vehicle sales and building permits are above the baseline, showing a positive rate of change (these sectors were in the negative territory prior to the 2007 recession). This was the beginning of the housing bubble, and although housing did fall some, it was one of the stronger sectors during that recession. For some reason, vehicle sales trended closely to housing sales during this time. One other difference is that industrial production and capacity utilization had very negative rates of change prior to this recession. Prior to the 2007 recession, the rates of change for these two indicators were positive.
That there are some differences between Grid 2 and Grid 4 is not a surprise. The economy changes over time and different sectors can be in different stages when a recession hits. Despite some difference, Grid 2 and Grid 4 have a remarkably similar pattern, including the position of the MoC.
Recession Range of MoC
As recessionary conditions develop,the expected path of the MoC can be predicted. Because the MoC positions in Grid 2 and Grid 4 are relatively close, this identifies the likely range of where the MoC will be when a recession is looming. On Grid 5 , which shows the current economy again (same plotting as Grid 1), the area in red shows were the MoC is likely to be just before a recession begins. Further, considering where the MoC was six months prior to the 2007 recession (Grid 3), this points to the path the MoC will take as the economy begins to slow and move towards a recession. The red arrow shows this track. This path towards a recession is somewhat obvious. As the economy weakens, more indicators move into the negative rate of change quad, pulling the MoC to the left side of the grid. The negative trends leads to more indicators moving below the baseline, pulling the MoC towards the baseline.
If you go back to see the other grids, keep in mind that Grid 1 and Grid 5 are scaled to -30%/+30% on the horizontal axis. Whereas, Grid 2, Grid 3, and Grid 4 are scaled to -50%/+50% on the horizontal axis.
As a next step, I’m going to map the indicators one-year and nine-months prior to the 2007 and 2001 recessions and hopefully see the period when the MoC moves into the negative rate of change grid. As I said earlier, that seems like a critical step towards a recession, maybe the point of no return.
One Lesson Learned
Over the past couple of years, as I have been following various indicators and tracking trends, I have realized that one or a few indicators alone don’t tell the full story about the economy. Yet, when one indicators changes significantly, pundits and headlines tend to focus there. The change might prove to be meaningful, but probably only if other key indicators are moving the same way. Mapping the above grids has reinforced that view. It is the sum of the parts that really matters and there is sufficient time to determine if a recession is brewing.
Notes and Comments
In case you are interested, here are some insights into the process I am using. Feedback is welcome. Also, at the end of the article is a list of sources for the economic indicators used here.
Adjusted Rate of Change
I have adjusted the rate of change to give every indicator a common-size rate of change. One problem with comparing indicators is that some have a small range of change and others a large range of change. To get a common-size rate of change for all indicators requires using each indicator’s unique range of change as the base to calculate the rate of change (Rate = Amount of Change / Base). For example, for an indicator with a 100% range of change, a 30% rate of change would easily be recognized as an effective 30%. But, the rate is only 30% because the base happens to be 100, such that 30 / 100 = 30%. On the other hand, for an indicator with a range of change that never exceeds 10%, a 3% change is not an effective 3%. In actuality, a 3% change with a 10% base would be a 30% rate of change (3 / 10 = 30%). By using the range of change as the base, this makes the rates of change proportional, as 30:100 is equal to 3:10, resulting in common-size rates of change.
In addition, to keep the range of change from getting skewed by outliers, I’ve eliminated range changes above or below the third standard deviation from the mean.
Composite Nine-Month Trend Versus Year-Over-Year Change
As I outlined in my last article, I am using a composite nine-month trend that weights the first three months the heaviest. I calculate the rate change over the most recent three months, over the most recent six months, and over the most recent nine months. The three-month trend is weighted at 60%, the six-month trend is weighted at 30%, and the nine-month trend is weighted at 10%. The weighted rate of change is then annualized.
Concern has rightly been expressed about seasonality influencing the trend because I am not using a year-over-year change. I feel that the weighted nine-month trend is more useful for my analysis for the following reasons.
- I want the mapping to be responsive to the recent changes of each indicator. I’m trying to forecast movement towards a recession and I want to capture the acceleration of any trends that signal a slowdown.
- When the economy begins to nose dive into a recession, the rate of change due to weakening conditions far outweighs the effects of seasonality.
- Seasonality mostly effects one axis on the grid, the rate of change. Movements towards the baseline stay within a tight range. When there is a significant shift in that range, it is due to more than seasonal factors.
- During a three- or four-month period, seasonal factors may work in favor of the economy, or against it. Capturing some of that influence is fine for my analysis.
- Most of the measures I use are seasonally adjusted, so any seasonal influence is already being moderated.
- Having a composite nine-month trend does temper the seasonality somewhat.
- Finally, the MoC is really the key indicator to follow. Because the MoC is an average of coordinates, the seasonal effects of one sector are likely to cancel out the seasonal effects of another.
My approach may result in some seasonal influence, but more importantly, it allows the most recent three months of data to have the greatest influence on how the grid is plotted, providing a more accurate snapshot of current economic conditions.
Establishing a baseline for some measures is easy because organizations that track certain data have established recession breakeven points. For example, ISM has breakeven points for both the manufacturing and non-manufacturing sectors. However, for some measures, especially those that grow over time, like corporate profits, establishing a baseline can be tricky. For most measures that don’t have an establish breakeven, I generally use the 15-year average, which seems to work for most measures. Sometimes there is an established pattern, like for corporate profits. For the last three recessions, the recession didn’t start until nonfinancial corporate profits dipped below the profit peak of the last business cycle. This relationship may be a bit tenuous, but for now, I’m using this as the baseline for corporate profits. I have also found when I adjust some measures for inflation or population growth, trends are more consistent, making it easier to establish a baseline. As I mentioned earlier, because the MoC was nearly at the baseline at the beginning of the 2001 and 2007 recessions, my baseline measures must be okay, at least when taken together. Still, there are a few indicators where I need to refine the baselines.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.