WARNING: unbalanced footnote start tag short code found.
If this warning is irrelevant, please disable the syntax validation feature in the dashboard under General settings > Footnote start and end short codes > Check for balanced shortcodes.
Unbalanced start tag short code found before:
“I-B).values), B.index, B.columns) #drop excess row and column del ib_inv[‘Households ; Final Consumption Expenditure’] ib_inv=ib_inv[:-1] #Calculate output multipliers iomult=ib_inv.sum() print(iomult) #calculate income multipliers incmult=pd.DataFrame(np.dot(x,i…”
As we write this, due to the Wuhan/Coronavirus the Australian Football League teams were playing matches in front of no spectators over a shortened season before the league was delayed entirely. The Western Australian Football League has been delayed until June, while junior training and games in Western Australia were also cancelled until June.
We don’t know where this will end, but we acknowledge footy is about to do it very tough for a while, as will many other sports and industries. While we economists tend to quote aggregated statistics, we should never forget it’s real people behind those figures.
We wanted to do a post that gave the reader an intuitive introduction into the world of input-output multipliers and to industry assistance evaluation in general. The opportunity to frame it within the AFL gives it a real-world feel.
The Wuhan/Coronavirus adds a necessary extra dimension to the discussion, one that if we’re honest we might not have considered otherwise.
We intend to follow up this post with further posts on multipliers, cost-benefit analysis and computable general equilibrium modelling. By the end of it you should have a good grasp of every tool a practical economist might need to analyse industry assistance proposals (and more). You will also know when to use each tool.
We’ll also post later on which tools are appropriate to evaluate policy proposals to recover from the Wuhan/Coronavirus virus.
Anyway, we spied an article a while ago on WA Today. which links to the Western Australian Football Commission site. The WAFC site contains an ACIL Allen report on the Economic and Social benefits of the AFL in Western Australia.
The report states:
Overall, ACIL Allen estimates that the WA football industry contributed $220.3 million to the WA economy in 2016-17. This is equivalent to 0.1 per cent of the State’s economy (Gross State Product) or 17.2 per cent of the Arts and Recreation Services sector.
The implied economic value added multiplier is 2.00 (the actual multiplier is 1.966), which means that for every dollar spent by the WA football industry in Western Australia, additional spending of $1.00 is generated across Western Australia’s economy.
ACIL Allen, “The Economic and Social Benefits of Club Based Football in Western Australia“, Report to the West Australian Football Commission, October 2018, pii
ACIL Allen’s report notes that the economic value added multiplier for regional mining was 1.43, so ‘Footy claims bigger dollar for dollar economic impact in WA than mining’.
As an aside, you shouldn’t compare a multiplier inflated figure with a non-multiplier inflated figure, as ACIL Allen appears to do with Arts and Recreation in the quote above, but that’s not the main point of this article.
The ACIL Allen report also examines social benefits to AFL but, although economists are often asked to examine social benefits, we’ll stick to the calculation of economic benefits in this article.
What does a bigger economic value added multiplier mean?
We’ll concentrate on what multipliers mean here. We don’t have the information to comment on the expenditure to which the multiplier is applied.
Let’s start with the elephant in the room, there are major problems with using I-O multipliers to measure industry assistance proposals.
I-O multipliers were popular around the turn of the century in Australia, but a considerable effort by economists (mostly) in Treasuries around the country discredited them and we at Practical Economics thought we had seen the last of them. But we guess what goes around comes around.
For those not familiar with I-O multipliers, they are calculated from an economy’s input-output tables. They are a measure of the flow-on impact of an additional $1 spent on the output from an industry under certain (very restrictive) assumptions.
They can measure the impact of expenditure of output, employment, income or value-added. ACIL Allen uses value-added multipliers here, which is the best multiplier to compare with the change in total size of the economy (we’ll do a post on differences between the various types later).
I-O multipliers are not the Keynesian multipliers you learned about in high school. Rather they are a measure of inter-industry linkages. An upcoming post will examine the difference in more detail.
They also measure economic impact (makes the economy bigger or not), rather than economic welfare (whether we make people better off or not). We’ll examine this issue in a later post, but it matters critically what question is being asked. ACIL Allen makes no welfare claims in its report, so impact analysis is appropriate.
A good summary of multipliers’ problems is contained in the Western Australian Treasury and Finance publication, starting on page 44. (The links on this have changed over the years, Google Western Australia Treasury Economic Research Articles if the links change again, it was published in 2002).
In fact the Australian Bureau of Statistics is so wary of their use that it no longer publishes them and it was tricky to find the numbers we needed to write this post.
Read the Western Australian Treasury and Finance paper to see all of the problems.
While not completely explicit, the ACIL Allen report contains two potential claims:
- a stand alone $1 spent on AFL generates $2 in the Western Australian economy, so we should find this $1 from any source and give it to AFL; or
- taking money from industries with low multipliers (e.g. regional mining via say a mineral royalty) and giving it to AFL would increase the size of the Western Australian economy (Gross State Product in this case).
The first claim roughly equates to the Keynesian concept that an increase in spending increases aggregate economic activity (usually Gross Domestic Product) greater than the original expenditure as income received from the expenditure is spent on a second round of goods and services, and so on.
As noted above I-O multipliers are not the same as Keynesian multipliers. In fact they’re just a clever piece of math.
However, in normal times, even if the figure did represent a Keynesian multiplier it shares similar weaknesses such as:
- the full multiplier would hold (if at all, this is controversial) if the $1 were not spent elsewhere in the economy, for example if the economy was in depression and people were hiding cash under the bed if it wasn’t spent on AFL. At most times an AFL club member would likely spend the $1 they spent on AFL of some other good or service (e.g. movies, a night out etc); and
- the full multiplier assumes that there is considerable excess productive capacity in the economy, so that new resources (workers, capital) can swing into action to generate the increase in production. Short of a depression or a very deep recession, much of the potential increase in output would be crowded out by resource constraints.
Having said this, we might not be in normal times. The economic impact of the Wuhan/Coronavirus looks extremely bad despite the best efforts of policy makers and central bankers. The key is to distinguish as to when these are problems and when they are not.
The big problem with AFL now is that people want to spend money but can’t (i.e. no games). It’s almost more of a supply restriction on top of a demand restriction, so we’re not sure a multiplier (I-O or Keynesian) could work in such a situation.
There is a case for government to help businesses stay solvent through what (we hope) is a temporary period of low demand, but I-O multipliers aren’t the tool we’d use to analyse this.
You just aren’t going to get any substantial flow on effects in the near term.
However, we don’t think that’s what the ACIL Allen (correctly) had in mind when it wrote the report in October 2018. More likely it meant the second claim, that a transfer of money from mining to AFL would increase the size of the economy.
In fact you might argue that even if the true AFL multiplier is way less than claimed, so might the true multiplier to regional mining. Perhaps the conclusion that AFL generates a bigger dollar for dollar impact holds?
Possibly, but it’s actually the wrong question.
We can and do take money from mining and give to other sectors of the economy. Transferring money to areas of social value not provided by the market is major role of government.
For mining in particular, we believe that the exhaustible nature of mineral resources gives those resources value even if they are still in the ground. We (Western Australia) charge mining companies mineral royalties on the resources they extract.
Without royalties, mining companies would gain the value of the un-mined resource and make super-normal profits over and above the return needed to invest in the project. In theory, because these profits are super-normal, imposing royalties does not affect miners’ decision to invest (the ‘correct’ amount of profit claw back was behind the whole resource super profits tax debate in Australia between 2010 and 2014).
So yes, we can, should and do take mining profits and distribute them to other sectors of the economy, including most directly government.
The real questions are whether:
- does AFL deserve more from current government transfers than it gets already; or
- should we take more from mining to give to the AFL?
On the first point, by the metric presented by the WAFC, there are industries with bigger value-added multipliers than the AFL.
We don’t know the exact input-output table used by ACIL Allen as, to our knowledge it hasn’t published all of its multipliers in one place. However, we decided to use the latest ABS Australian tables (2016-17) to calculate our own multipliers, with the method as outlined by the ABS. The Python code for our estimates is shown at the end of this post.
There is no AFL industry in the national I-O tables, but the industry to which the AFL belongs, Sport and Recreation, has a value-added multiplier of 1.99 – close to the ACILAllen estimate.
(As an aside, value added multipliers are much bigger than we recall from the early 2000s, we must look into this)
But, to name a few, Public Administration and Regulatory Services has a value-added multiplier of 2.16, Employment, Travel Agency and Other Administrative Services 2.22, Public Order and Safety 2.21, Primary and Secondary Education Services Residential Care 2.30, Social Assistance Services a whopping 2.34, Technical, Vocational and Tertiary Education Services 2.24, Clothing Manufacturing 2.12, Health Care Services 2.15 and so on.
Additionally, buried in the report (p15) is that ACIL Allen’s own value-added multiplier for a mineral sands project in the Peel region is 2.33 (to be fair the WA Today article did include this table).
Why does the AFL deserve funding ahead of these noble industries? On economic impact grounds why would we build a new stadium rather than more schools and hospitals?
The economic ‘kick’ from post-virus stimulus policies will be a key discussion at the right time, but we’d argue I-O multipliers aren’t the right tool for this.
On the second point of royalty increases, royalties should be set at a level so that the Western Australian public receives as much benefit from the resource they own as possible, but not at a level where mining companies reduce investment in their mining sector (this is never an exact number of course, we just don’t have that level of knowledge).
One or two cancelled resource projects would dwarf the economic impact AFL.
So we should set royalties at their appropriate level and then decide where best to spend the money.
Overall, any claims to increased funding to AFL activities are shaky at best.
If input-output multipliers are so bad, they why were they so widely used in the past? The answer is that they are simple and they give a nice big number to inflate the importance of an industry. Most commonly such claims were associated with requests for government subsidies on the grounds that the industry was so important.
People also have the Keynesian multiplier concept in their head and so latch onto the word without knowing its full meaning.
This doesn’t mean money spent on footy does not contribute to the economy – it does. It’s likely beneficial at a consumer level, as it’s where a substantial number of people prefer to spend their money (although you’d need to untangle all of the government subsidies the sport receives for a definitive answer).
The macro-economy might be better off because it enables many high paid (i.e. productive) jobs to be located in the state.
It has economic benefits.
Just not $2 for every $1 spent, at least most of the time.
Python Code
Below we provide our Python code for calculation of Australian multipliers from ABS 2016-17 I-O tables. Our philosophy is to publish all of our calculations where practical, which means you get to see if we’re wrong.
However, we haven’t published the multipliers estimates here. The ABS doesn’t publish them on the grounds that people with no technical knowledge misuse them badly. An old colleague of ours used to say you didn’t deserve to use multipliers if you couldn’t calculate them for yourself. We’re lowering the bar here to: if you can’t copy, paste and run a bit a Python code you don’t deserve to use them!
Please feel free to check the code and let us know if you have any comments.
To install Python see here.
import pandas as pd
import numpy as np
#Download Data from ABS website and organise
url=’https://www.abs.gov.au/AUSSTATS/subscriber.nsf/log?openagent&520905500108.xls&5209.0.55.001&Data%20Cubes&2C218AF595C8BFE1CA25843B00151D2A&0&2016-17&19.07.2019&Latest’
IOTable = pd.read_excel(url,’Table 8′)
IOTable=IOTable.set_index([‘USE’])
IOTable.index.name=’Industry’
del IOTable[‘Unnamed: 0’]
new_header = IOTable.iloc[0] #grab the first row for the header
IOTable = IOTable[1:] #take the data less the header row
IOTable.columns = new_header
IOTable = IOTable[~IOTable[‘Sheep, Grains, Beef and Dairy Cattle’].isnull()]
#Drop rows unnecessary to B* matrix calculation
IOTable_m=pd.DataFrame(IOTable)
IOTable_m=IOTable_m.drop(‘Total Intermediate Use’)
IOTable_m=IOTable_m.drop(‘Gross operating surplus & mixed income’)
IOTable_m=IOTable_m.drop(‘Taxes less subsidies on products’)
IOTable_m=IOTable_m.drop(‘Other taxes less subsidies on production’)
IOTable_m=IOTable_m.drop(‘Complementary imports’)
IOTable_m=IOTable_m.drop(‘Competing imports’)
IOTable_m=IOTable_m.drop(‘Total uses’)
#Get value added vector for later multiplier calculation
y=IOTable_m.T[‘Value Added’]
y=y[:-10]
z=IOTable_m.T[‘Australian Production’]
z=z[:-10]
a=y/z
#Now drop columns unnecessary to B* matrix
IOTable_m=IOTable_m.drop(‘Value Added’)
del IOTable_m[‘Total Industry Uses’]
del IOTable_m[‘General Government ; Final Consumption Expenditure’]
del IOTable_m[‘Public Corporations ; Gross Fixed Capital Formation’]
del IOTable_m[‘Private ; Gross Fixed Capital Formation’]
del IOTable_m[‘General Government ; Gross Fixed Capital Formation’]
del IOTable_m[‘Changes in Inventories’]
del IOTable_m[‘Exports of Goods and Services’]
del IOTable_m[‘Final Uses\n(Q1 to Q7)’]
del IOTable_m[‘Total Supply’]
io=IOTable_m.T
io=io.replace(‘n.p.’, 0)
#create B matrix of production coefficient shares
B=io.div(io[‘Australian Production’],axis=0)
del B[‘Australian Production’]
B=B.T
#Create income vector
x= B.T[‘Compensation of employees’][:-1]
#Create B* matrix (calculating only total multipliers here)
I=pd.DataFrame(np.eye(len(B),dtype=float))
I=I.set_index(B.index)
I.columns=B.columns.values
ib_inv = pd.DataFrame(np.linalg.pinv((I-B).values), B.index, B.columns)
#drop excess row and column
del ib_inv[‘Households ; Final Consumption Expenditure’]
ib_inv=ib_inv[:-1]
#Calculate output multipliers
iomult=ib_inv.sum()
print(iomult)
#calculate income multipliers
incmult=pd.DataFrame(np.dot(x,ib_inv),iomult.index)
print(incmult)
#calculate value-added multipliers
vacmult=pd.DataFrame(np.dot(a,ib_inv) ,iomult.index)
print(vacmult)