CHANGES IN U.S. MANUFACTURING

Aside from the persistent and important question of manufacturing job losses in the U.S., and whether these losses can be regained or at least stopped, it may be instructive to see the changes in the manufacturing industry's structure over the last 20 years or so. Changes in the structure, regardless of the job losses, have an impact on the average wages paid in the sector and, thus, impact the incomes of U.S. consumers.

As I pointed out in an earlier post in this blog, manufacturing shipments have been relatively robust. Overall they have followed a positive trend, of course allowing for declines associated with economic recessions. Since 1992 shipments have doubled, translating into a 4.4% annual growth rate.
However, during this same period prices measured by the GDP Implicit Price Deflator rose by slightly over 50%; that is a 2.4% annual rate. After deflating the shipments data we find that the 4.4% annual growth is only 1.3% in real terms. Naturally, applying such a broad price measure does not give an accurate measure because price inflation varies among industries. Case in point is the sharp drop in petroleum and natural gas over the last couple of years, a decline that did not result in similar price reductions in other industries that depend in oil products.

Changes in Industry Structure
Since 1992 there have been several important changes in the structure of manufacturing, as one would expect. An economy is not a static entity but, rather, is one in which change is prevalent. At the micro level, old factories close and new factories spring up, production in individual factories goes up and down depending on the whims of consumers who may want more or less of the products made by those factories. At the macro level, we see new industries being born making products that nobody may have thought of up to that point, event though many make claims to it, or we see whole industries lose significance or disappear altogether.

The chart to the right compares 21 broad manufacturing sectors in 1992 and 2015, where we use shipments data through September for each of the years. The charts allow us to see two types of changes. One is in the percentage of manufacturing accounted by each sector, such as the 1.4 percentage points gained by Transportation Equipment (that includes autos, trucks, aircrafts, etc.) or by the Food sector.
The second type of change is the relative ranking of the various sectors within manufacturing- these changes are indicated by the green and red arrows and they highlight significant changes. Such is the case of Petroleum and Coal that, unsurprisingly, jumped to number four with 5.1 percentage point increase in share. At the same time we see sectors such as Computer and Electronics that fell three places to number seven with a 3.4 percentage point decline. Although I am not investigating causes for these declines, one can surmise that this reflects production shifted to overseas locations for cost considerations.
The good thing is that four out of the top five manufacturing sectors, that account for nearly half of manufacturing shipments, are also among the ones with the highest hourly earnings; the exception is Food Products Manufacturing with workers in this sector earning 24% below the average hourly manufacturing wage.

Manufacturing Losing Ground
Compared to other industries within the U.S. economy we find that manufacturing is not keeping up in general with overall growth in the economy. In terms of total output, manufacturing accounted for over one quarter of the output of all private industries in 1997 (the earliest year for which we have data). By last year, the share of manufacturing had fallen to just over a fifth of gross output.
Some of the decline in manufacturing is due to the long-term shift towards increased reliance on services. For instance, health services is an area where usage has increased, as shown by the two percentage points increase in the charts to the right. An given Obamacare's mandates, we should expect health's share of gross output to increase further in the next few years.

HOUSING STARTS AND EMPLOYMENT

Is there a relationship or link between new jobs and housing starts? Economic reasoning would logically lead you to answer positively- the greater number of people who get a job, the more likely some of them would purchase a new home. Of course this is all contingent on other factors such as the availability of vacant housing, ability to get a mortgage, and a growing population, to name a few. That is, any direct link between new jobs and new housing is constrained by demographic and other factors.

However, the search for a way of predicting the course of new housing starts led some economists, a few years ago, to posit a fixed or constant relationship between new jobs and housing starts. They sought and came up with a constant value that could be used as a rule of thumb to calculate how many new houses we should expect given the growth in employment. Today I examine whether such a constant does in fact exist or can be calculated meaningfully. I do this analysis at two levels; first at the national level looking at aggregate data that we normally see on a day to day, and then using the less commonly seen metropolitan area. Does such a relationship hold?

U.S. New Jobs and Housing Starts
At the national level we find that the ratio of new jobs to starts hovers somewhere between zero and three, ignoring those periods when employment falls in negative territory. A simple mathematical average using data from 1960 on, a calculation that one can always do with numbers even if the result is totally meaningless, shows the ratio for the U.S. equal to 1.1. Taking this number as a rule would mean that a new job translates into a slightly more than a new housing start. Applying this ratio to the number of new jobs between 2010 and 2014 results in over eight million housing starts driven by the new jobs created. But in fact, over this five year period, there were 3.9 million new houses started, and not eight million as the ratio suggests.

In reality, as a visual inspection of the graph to the right clearly shows, there is no stable ratio value. The blue line is the annual ratio of new jobs to housing starts going back to 1960. The red line reflects the 1.1 average over all those years (I excluded the years since 2008.) The way the blue line fluctuates around the average red line shows that the average does not carry much predictive power. In many years it underestimates the ratio and, conversely for others the ratio is overestimated; the size of the discrepancy is of an order of magnitude of more than two. One can only conclude that if indeed employment growth leads to new housing construction, we can't say with any degree of confidence how large or small the impact will be; that is, how many housing starts we should expect from employment growth alone.

Is there a ratio for MSAs?
We find that the relationship new jobs to starts is even more tenuous at the local level. We examined data for the ten largest metropolitan areas in the U.S. based upon the number of housing permits; we are using housing permits instead of starts, since the latter are not readily available for metropolitan areas.

A review of these data, graphed on the right, shows that there is not a consistent pattern for these metropolitan areas. In fact, the opposite seems to be true, there is great variation both within and between metro areas. Following any single line, for instance looking at the top line that is Los Angeles, we can see that the ratio ranges from 5 to 11- a huge difference. Also looking at the values for a given year, say 2011, we can see large differences between metro areas. The ratio can range from a low of around 3 for Atlanta to nearly 10 for New York. Thus there is not an accurate rule of thumb that we can use given there is not a stable value that can be used for any of these metro areas.

All of this simply shows that suggesting a specific number of housing starts given the growth in new jobs is close to economic nonsense. It is correct to say that more jobs will likely lead to more housing starts, the same that it will lead to more automobile sales or any other consumer product. People work because they want to buy stuff. However to say that X number of new jobs will produce exactlyY housing starts is a totally false statement.

WATCH OUT FOR MISLEADING DATA!

The Social Security Administration, the agency managing the largest Ponzi scheme in the world that some would call the "mother of all Ponzi schemes", has just released data on the wage and income compensation paid to workers in 2014. The data is somewhat interesting, mostly from a curiosity angle since its analytical value is minimal. Nonetheless, I think it's worthwhile reviewing it, and pointing to its limitations since the data may easily lead the unwary to incorrect conclusions. In fact, I've seen a couple of references to this data drawing incorrect comparison to poverty levels.

The data shows the total wage income of individuals that is subject to social security payments. You can look at the raw data for last year here: https://www.ssa.gov/cgi-bin/netcomp.cgi?year=2014. Data for previous years is also available in the same web page.

The chart to the right shows on the horizontal axis the wages and income paid to all workers in the U.S. in 2014, in ranges of five thousand dollars up to those who earned $200,000. From that point the data are grouped into $50,000 ranges up to the million dollar mark, etc. On the vertical axis is the number of workers falling within each range. We can see the first point on the left represents 22.5 million workers, earning each under $5,000 a year; this is just over 14% of the workers. Wow, is the data telling me that one out of seven workers in the U.S. made less than $5,000 last year? Things are pretty bad then.

Well, not really. This is where the data is misleading if it's interpreted wrongly. First of all, the data includes all workers; teenagers working part time in the Summer, retired workers who also work part time in temporary jobs, housewives and others who take temporary jobs around the Christmas season, etc. Secondly, many of these workers are not the primary income earners in a household, they may have a spouse who is the main income earner in the family. That is, this is not the normal household or family income data that is more meaningful for analysis. For instance, the median household income is currently about $53,000, but using the Social Security data we find that the median compensation per person is about $44,500- the difference is made by other working persons living in the same household who presumably contribute to the household's economic wellbeing.

Comparisons to other data, such as poverty levels for instance, should not be made using these data. Poverty levels are usually meaningful only in terms of the number of persons who live within a family, while the social security data above does not provide a hint whether the person is living alone or with other members. If one were to use these data to estimate the number of people who are below the poverty level, we get that about 40.5 million individuals fall below the poverty line of $11,670 for one individual- this is 25% of the wage earners. But this is the wrong conclusion, as I said, because that individual may or may not be part of a larger family.


However an interesting comparison is to see how the data changes over time. The chart to the right displays the changes in the number of wage earners over the ten years ended in 2014. The horizontal axis displays data for individuals making less than $100 thousand, split again in groups of $5,000 each. The bars highlight changes over two 5-year periods; one is the change from 2004 to 2009 shown in the blue bars, and from 2009 to 2014 in the red bars.
It is immediately apparent that the number of individuals earning less than $35,000 fell over the ten-year period; it fell by nearly 8.4 million workers (although it's hard to discern that from the graph alone.) The graph suggests that most of that change occurred between 2004 and 2009, when in fact the number of individuals making less than $35,000 fell by 6.9 million.

Another caution with the data can be inferred from the last paragraph and chart. Yes, the number of workers earning less than $35,000 fell sharply; this is not necessarily because their incomes improved but because many of them lost their jobs as a consequence of the 2008-2009 recession.