output per worker in construction sectors
I'm trying to understand how the output per worker is derived in the construction sector and reconcile it with other ways of estimating it.
From the V2 documentation and just looking at the numbers, it appears that your estimates of employment are based on BEA's estimates. For New Jersey, which is what I'm looking at, you have construction employment of 265,630, compared with BEA's 251,786 in 2008. However, the 2007 Economic Census gives a total of 178,290, the Census' County Business Patterns in 2008 gives 172,004, and the BLS' QCEW gives 166,183. I assume that Differences between the latter three can be explained by timing or minor methodological differences. But your number and BEA's, which are relatively close, are quite different from these last three. What accounts for the difference?
Also, the V2 documentation mentions that "state level construction values are combined with output-per-worker estimates and earnings-per-worker estimates derived from the current national input-output study to form a set of employment and earnings estimates for each state." Has the methodology changed? The national model has output per worker of $138,121 in sector 36, compared with $154,436 in New Jersey.
Thanks.
-
Census, CBP and QCEW are wage and salary employment only. BEA and IMPLAN are wage and salary plus self employment. By BEA, I assume you are referring to the REIS numbers, there are two things working for the differences between IMPLAN and BEA. First, BEA revises the previous five years of estimates every year, so the number you are using for comparison is likely to be different then the numbers we used. Second, we use I-O redefinitions on our industry data. One of the redefinitions is to remove new construction/investment activities (transmission line construction, computer software development) from the industry doing such work "in-house" and moving it to the appropriate sector. This works to increase the new construction numbers. The methodology is similar. Your example indicates that NJ has higher earnings per worker. We equate higher output per worker to higher earnings per worker. -
Thanks Doug. I took a look at the Nonemployer Statistics. The self-employment statistics would seem to push down the output per worker pretty considerably. For sector 23 for New Jersey, receipts per establishment average about $69,000, well below the output per worker for wage and salary employees. Am I interpreting this correctly? If this is the case, how should I estimate the direct impacts of a construction project that I know is going to hire a large construction firm? Thanks. -
I believe you are saying that Census has the average output per proprietor for "non-employer" establishments as being less than output per worker for the Census establishments with employees. I have no problem with that. The direct impact would be the value of the structure being built (net of land value). A general discussion of construction projects can be found in this thread: http://implan.com/V4/index.php?option=com_kunena&func=view&catid=80&id=5750&Itemid=35#5750 -
Doug, I guess my concern is that applying construction spending (the portion that goes for structures, not land or FFE or A&E) to one of the construction sectors would overstate the direct job impact in our case, where we know that a large construction firm would be doing the work. Would you suggest calculating the direct jobs off-model in this case and use the model to get the indirect and induced impacts? Thanks. -
Finally, I understand the question. There are two things to keep in mind if calculating the direct employment separately: 1) It is the labor income that drives the induced effect - so is more important than direct employment. 2) Construction sub-contractors in I-O are part of the direct effect. Their employment is direct (rather than indirect). As a sub-contractor they get their contracted portion of the direct. -
OK, so here's a conceptual methodology: INPUT: construction industry spending 1. Calculate output per worker based on the Economic Census figures. 2. Calculate labor income per worker. The Economic Census provides payroll and benefits. I'm not sure how I can derive proprietor's income only for firms and not get self-employed proprietor's income mixed in. 3. Apply these factors to the industry spending number to get direct jobs and then labor income. 4. Feed labor income into IMPLAN to get the indirect and induced impacts. -> Any ideas on how to determine the proprietor's income component of labor income, without getting self-employed workers' proprietor's income? -> Regarding point #2 in your last response, is the output per worker that I calculate via step 1 going to miss the subcontractors? -> Am I risking a geographic mismatch for the labor income? I would be feeding labor income into the model as if it were all going to New Jersey residents, whereas some of it might go to neighboring states. Thanks. -
Proprietor Income = Self-employment income. I would assume that Proprietors who also have wage and salary employees would generally have higher income that proprietors with no employees - but I don't have data for that. You may have to just live with the ratios you derive which will include both. Census output per worker will be for all levels of sub-contracting. Applying the ratios to the total value of the structure should be appropriate to capture the employment of the primary as well as the subcontractors. I don't know enough about the project or the region regarding in-commuting. It is possible if not likely that if the primary contractor is non-local that much of the administrative work and the "specialty" contractors may be brought in. I would think that most of the general subcontracting would be local. If you can specify a % of labor that is local/non-local, you can use that percentage to re-specify the employment compensation and proprietor income in the event line to reflect only that is local. -
Doug, This sounds good. I had thought that we would calculate direct jobs and labor income outside IMPLAN and then feed labor income to the model. Could we get the same result by just changing the output and labor income per worker in the model for the construction sectors? Thanks. -
Yes. The simplest method would be to edit the impact event (see http://implan.com/V4/index.php?option=com_content&view=article&id=568:568&catid=237:KB20 ). The software applies the Sales Value to the type I multipliers to generate indirect effects. It then takes the direct employee compensation and proprietor income specified in the event along with the indirect labor income to generate the induced effects. -
Doug, I want to review what I'm doing to make sure it makes sense: We've calculated output per worker for the construction sector, based on Economic Census data (thus excluding the self-employed workers). For the numerator in that calculation, I'm using the value of construction work less the cost of work subcontracted out to others. We also have our own estimates of wages + benefits per worker, based on local union wage schedules (for the trade workers) and local BLS occupational employment stats (for the non-trade workers). Labor income accounts for roughly 40-60% of output. After deducting compensation from output, we had to estimate the portion of the remainder to apply to indirect spending. Based on the Economic Census, we estimate about 75% of the remainder goes to indirect spending, with the remaining 25% leaking out to capital expenses, land, profit, and miscellaneous taxes. We want to use some of the more detailed construction sectors that existed in the 509 sector scheme, so we created Industry Change activities based on the 509 to 440 spending patterns library. To estimate the impact of a million dollars in spending in an industry, we first divide by output per worker to get the direct jobs. Direct jobs x compensation per worker is fed into the model as labor income. Subtracting labor income from the output, we take 75% of the remainder (the portion that we estimate to be indirect spending) and apply that as an Industry Change to the spending pattern activity mentioned above. The sum of the Labor Income and Industry Change impacts becomes our indirect+induced impact. Does this sound reasonable? Thanks.
Please sign in to leave a comment.
Comments
11 comments