Should an Employment Event be used alongside a Contribution Analysis?

Hello:

So you're saying that if I know something like sector 511 Food and Drink is experiencing 100% unemployment that I should put -100% in Industry Contribution, as opposed to minus all employed and output?

Now what if the employment of an industry is changing but not the output?

Hello -

Using an Industry Contribution Analysis, you would model the loss in Output. So yes, 100% closure of the Industry 511 would be 100% in an Industry Contribution Analysis Event. Note using this will give you an associated Direct Output, but you can zero that out if you want to make the assumption there is no change in Output even if there is no one working in that Industry.

Okay, thank you for those answers.

I guess to just get it clear:

Problem 1:

I'm attempting to show the effects of non-essential industry closures on all industries concerned. Let's say 3 months. Should approach this by Contribution Analysis and inputting these industries at -25% (representing 3 months of the year)?

This will answer the following questions, correct:

• How many employees will this affect (in Direct, Indirect, Induced)
• How many dollars in output lost will this result in (in Direct, Indirect, Induced)

Problem 2:

I'm attempting to show the effects of non-essential industry closures who have furloughed employees but still have output. Let's say 3 months again. Should I still approach this with Contribution Analysis at -25% (representing 3 months of the year) or should it be broken into an output event and employment event?

Problem 3:

If these analyses are less cut and dry than 3 months offline and then 100% back... how would I measure these industries stepped at say 3 months no output, 3 months at half output, 3 months at 3/4s output.

I'm sorry to ply you with so many questions.

Solution 1:

You nailed it.

Solution 2: 

These companies likely have some Output, but not to the pre Covid 19 levels. How you want to examine the loss in Output is up to you to determine. The 25% will show the 3 month total loss in the Industries. You may want to tweak that number to represent that there is some Output (and employment) still happening.

Solution 3: 

I recommend using the average Output for the year with your assumptions of when there will by 100%, 0%, 50%, and 75% activity in each quarter.

I'll preface this by saying I've only had to run I/O analyses before for new establishments or jobs created/lost, so that you for helping.

So I have run a Contribution Analysis at -25% for 'Large Group' of industries and one for 'Smaller Group' of service-only .. but when comparing the results I see that there is no difference (and many times no Indirect or Induced) between the results on an industry level.

For example, in my Large Group analysis it shows -893 employees, but also at within my Small Group analysis it shows -893. I would assume that the more industries that see -25% in my Large Group, that there would be sectors that see higher loss of Indirect and Induced due to the loss of consumer spending from more of the communities jobs being eliminated for 25% of the year..

When you analyze an Industry Contribution Analysis with multiple Industries, the buybacks to each of those Industries will be restricted. Therefore, you won't have any associated Indirect or Induced Effects in any of the Industries you included in your Events.

Okay, but if I am looking to see the Indirect and Induced effects as well, while running the sweeping lockdown effects, am I better off to run an Employment Event and subtract the 1/4 of total employment for all the affected lockdown industries instead of running the Contribution analysis across so many sectors (100+) .. I fear that losing the buyback effect is limiting the real outcome of reduced worker spending.

If you run a standard Industry Event of a loss of 25% in say, full-service restaurants, you will see the Indirect and Induced effects that will also affect full-service restaurants. This may overestimate the actual losses that are being seen. If you want to assume an initial loss of 25% that will have additional losses, then you can use an Industry Event. 

If you know that there is a 25% of 100+ industries and no more, then you want to stick with Industry Contribution Analysis. 

Okay, I may want to run employment events then, as I do wish to see the indirect and induced effects.

However, I am basing this off of our MSA unemployment data by month, so I do have a question there: if I run a negative employment event, it applies towards the whole year within the model correct? So if I were to say -100 employment in sector 511, it would calculate the output lost for that whole year (and its indirect/induced). If I were to take the results of this impact and then divide those results by 25%, would that reflect an estimation of the negative employment event over just 3 months (in output dollars)?

You got it!

Okay, last question (I hope!)

I've now run my impacts, two of them, one for all of my industries, and one for just a subset of those.

The results say that for the larger analysis(let's say 250 industries) that the employment impact is less than the employment impact of the smaller analysis (say 150 industries) for some industries..

I will say that the total employment change was larger for the larger analysis, and smaller for the smaller.

So then why would some industries experience a 'lighter' effect from a 'harsher' economy-wide employment event?

I've run a constant percentage within my own designated super-sectors (say 7% for everything I deem Manufacturing, etc) but all impacts were as reductions, none positive.

I would think that with the 'larger' analysis that it all around meant less money circulating and so every industry affected would be affected more negatively. Unless some have relationships wherein the inverse is true.

Hello - 

It's really hard to make a guess as to what's going on without seeing what you ran. My first inclination is that you are seeing Aggregation Bias. You will generally see larger analyses having larger impacts, but that isn't always the case.