How do we handle the special commodity Scrap?

[WesIngwersen]

'Scrap', S00401, is a special commodity that is produced as a by-product of industry but not classified by type in the BEA data.

more from BEA (2009):
"Scrap sold by final users is not part of the gross output of the economy, but it is part of the supply of goods available. All scrap originating in final uses is sold to intermediate. Examples include sales of used aluminum cans by persons to recyclers and sales of scrapped vehicles to junk dealers."

Scrap was adjusted following the BEA procedure to remove it as a commodity from USEEIO v1.
Scrap was included in USEEIO v2 as a commodity.
Scrap is not included in CEDA.

Resolved: Make the scrap adjustment for v1. Consider adding further specificity in the future and using additional data on country-specific scrap input percentages to customize a direct requirement for primary inputs when using it for structural reflection.

[MoLi7]

In CEDA, Make table is scrap-adjusted first then used to derive V_norm (derive_2017_Vnorm_scrap_corrected in bedrock repo)

[jvendries]

Scrap adjustment is a type of reallocation, similar to what we discussed in #3 for the after redefinition tables. We should remain consistent with the decision there and include the scrap adjustment if we use the after redefinition tables.

[WesIngwersen]

The scrap adjustment is different from redefinititions because it does not move the production of (and associated requirements to produce) scrap to a primary industry. A non-scrap ratio is created representing non-scrap output fraction of each industry, and then each industries non-scrap commodity output is inflated by it in the market shares matrix, so their shares of non-scrap commodities add to 1.

However I agree that for each industry that produced scrap it results in a cleaner profile of production because they are no longer producing it, and producing more of their respective primary products or secondary product relatively, so in that sense it is like the redefinitions.

But I would suggest more analysis is needed because their is both information loss and a potential skewing of direct requirements for scrap using sectors as a result of its removal from the use table.

[WesIngwersen]

Here are the industries with the highest uses of scrap relative to their total intermediate uses.

Note that the steel and secondary aluminum industries real stick out with ~30%.
Paper/cardboard and glass industries are also heavy users of scrap, but the market value of used paper and glass is so low that it makes up a very small percentage of cost of intermediates.

Industry	Industry Name	Use of S00401/US relative to Total Intermediate
331110/US	Iron and steel mills and ferroalloy manufacturing	0.30
331314/US	Secondary smelting and alloying of aluminum	0.30
331510/US	Ferrous metal foundries	0.26
331200/US	Steel product manufacturing from purchased steel	0.07
322130/US	Paperboard mills	0.05
322230/US	Stationery product manufacturing	0.04
33131B/US	Aluminum product manufacturing from purchased aluminum	0.04
322299/US	All other converted paper product manufacturing	0.04
33291A/US	Valve and fittings other than plumbing	0.04
331420/US	Copper rolling, drawing, extruding and alloying	0.04
331490/US	Nonferrous metal (except copper and aluminum) rolling, drawing, extruding and alloying	0.04
331520/US	Nonferrous metal foundries	0.03
322110/US	Pulp mills	0.02
322120/US	Paper mills	0.02
327200/US	Glass and glass product manufacturing	0.01

[WesIngwersen]

Similar results are found in comparing the A matrix without and with the scrap adjustment. Missing scrap a few commodities have considerably fewer total intermediate requirements

Commodity	Sum Intermediate Reqs	Sum Intermediate Reqs Scrap Adj	Rel_Diff
331110/US	0.75	0.55	0.36
331510/US	0.59	0.44	0.34
331313/US	0.83	0.66	0.25
331200/US	0.67	0.63	0.06
322130/US	0.67	0.64	0.05
33291A/US	0.55	0.53	0.04

[WesIngwersen]

Expanding on comparison... the A matrix coefficients do not show much change.

The one point that sticks out at about [.22,.22] is the input of 331110 to 332119 which changes by 0.01.

[WesIngwersen]

Expanding the comparison, the N matrix changes are more significant.

The absolute relative changes are over 10% for many commodities

	N	N_adj	N_rel_abs_change
33131B/US	0.628169	0.937976	0.493190884
331420/US	0.384086	0.546954	0.424042838
332430/US	0.40167	0.559969	0.394103738
334210/US	0.108799	0.066752	0.386461209
325412/US	0.176655	0.115047	0.348751732
334515/US	0.10298	0.070804	0.312450228
334220/US	0.102666	0.071316	0.305352988
324110/US	1.574327	1.133948	0.27972489
334511/US	0.112333	0.081459	0.27484181
331410/US	0.513971	0.651333	0.267256685
336992/US	0.411185	0.520177	0.265070091
334112/US	0.084591	0.062241	0.264213416
334514/US	0.136391	0.100428	0.263676932
335911/US	0.478224	0.600032	0.254710972
33451A/US	0.121931	0.092879	0.238265323
325414/US	0.276447	0.211363	0.235432919
336411/US	0.201331	0.155491	0.227685271
424200/US	0.14833	0.115462	0.221585614
325411/US	0.352613	0.275349	0.21911928
334118/US	0.191131	0.150674	0.211667678
334300/US	0.113817	0.089997	0.20927872
334512/US	0.092489	0.073368	0.206737361
424700/US	0.159687	0.127999	0.198438222
334513/US	0.083548	0.067041	0.197569209
331490/US	0.574128	0.68456	0.192348471
339116/US	0.162539	0.193608	0.191146982
336412/US	0.213349	0.174211	0.183444932
311920/US	0.523784	0.615418	0.174944259
5241XX/US	0.046251	0.038169	0.174755054
334290/US	0.095226	0.079565	0.164463116
334111/US	0.055011	0.046205	0.160089943
335920/US	0.384434	0.445166	0.157979364
331520/US	0.487248	0.562337	0.154107391
326160/US	0.840494	0.966934	0.150435483
336414/US	0.187067	0.159382	0.14799177
327991/US	0.360137	0.307481	0.146210883
517210/US	0.112891	0.09653	0.144926618
811200/US	0.114294	0.097956	0.142949003
517110/US	0.098837	0.084815	0.141867853
334516/US	0.130721	0.112266	0.141175959
336413/US	0.211267	0.181562	0.140605909
517A00/US	0.100436	0.086397	0.139777886
512100/US	0.062869	0.054392	0.134823873
233262/US	0.305395	0.265088	0.131981007
S00500/US	0.146937	0.127906	0.129517919
311990/US	0.765046	0.861767	0.126424527
324190/US	1.369537	1.196988	0.125990873
711500/US	0.043851	0.038344	0.125582719
335312/US	0.270316	0.236434	0.125345148
336320/US	0.328365	0.287295	0.125074466
326110/US	0.810057	0.907278	0.120017542
524113/US	0.062319	0.054848	0.119888717
518200/US	0.10636	0.093625	0.119735951
335999/US	0.22994	0.202427	0.11965394
334418/US	0.103946	0.091525	0.119494674
326120/US	0.741199	0.828483	0.117761157
333993/US	0.197182	0.17432	0.115943827
233210/US	0.254903	0.227189	0.108724194
54151A/US	0.090207	0.080426	0.10843057
515200/US	0.110887	0.099007	0.107143288
33291A/US	0.291221	0.260228	0.106424891
322220/US	0.542461	0.599621	0.105372
326140/US	0.792519	0.87562	0.104857231
811400/US	0.165122	0.147968	0.103888826
485000/US	0.579367	0.519294	0.10368742
335314/US	0.21827	0.195667	0.103556496
611B00/US	0.139674	0.125232	0.10339656
813A00/US	0.079753	0.071525	0.103167764
331510/US	0.506861	0.45463	0.103048983
314110/US	0.612409	0.67448	0.101355495
333912/US	0.286314	0.257411	0.100948498
33399B/US	0.317319	0.28531	0.10087369

[MoLi7]

BEA's Concepts and Methods of the US IO Accounts, APPENDIX TO CHAPTER 12 provides clear guidance on adjusting for scrap and then deriving Direct Requirements and Total Requirements after removing scrap from the I-O tables. This is the appropriate and defensible approach, and we should adopt it.

CEDA has already implemented the scrap-adjustment step in derive_2017_Vnorm_scrap_corrected. We can use that implementation as our starting point.

[WesIngwersen]

Before this decision is made, I recommend we perform these two tasks and incorporate the results here:

1. Determine how Used goods and Scrap are handled in non-US IO table #55 Determine how Used goods and Scrap are handled in non-US IO table. This needs to be understood because the scrap and used goods are substituting for a primary material. The mix of scrap and primary material can be very technology and country-specific.
   Claude found good data on this mix.
   https://claude.ai/share/2e195c57-13d9-4166-89e2-2b678fcbf9ce
   When you are removing scrap, following from this statistic that 69% of steel production is from scrap in the U.S., removing scrap means that 69% of the raw materials are not included in the recipe. This is OK for the U.S. if we treat scrap like a secondary material and the rule that we want to apply in the model is a cut-off approach where the use of a secondary materials carries no burden, but this is problematic if the A matrix is used in structural reflection to reflect a recipe for other countries, because it automatically implies that country is also displacing 69% of its primary material requirements with scrap.

2. Can model be validated using useeior validation after scrap adjustment? #56 Can model be validated using useeior validation after scrap adjustment?
   These validation procedures assure balance and proper accounting. We need to make sure they can be properly used or adapted after the scrap adjustment.

[WesIngwersen]

Validation tests passed for the scrap adjusted models with the same frequency as the standard models.
From
m_adj is the scrap adjusted model, m is the standard model. The useeior built-in "USEEIOv2.3-GHG" model specs is used.

methods/53-compareAbeforeandafterscrapadjustment/EvalScrapAdjustment.R

Lines 197 to 249 in f92e88f

	## Try to Validate adjusted model
	# Test if adjusted model output can be recalculated
	v <- compareOutputandLeontiefXDemand(m_adj, use_domestic=FALSE)
	v$Failure
	# rownames variable
	# 399 S00402/US V1
	# 400 S00300/US V1
	# Test if model output can be recalculated
	v <- compareOutputandLeontiefXDemand(m, use_domestic=FALSE)
	v$Failure
	# rownames variable
	# 399 S00402/US V1
	# 400 S00300/US V1
	# Test if adjusted model domestic output can be recalculated
	v <- compareOutputandLeontiefXDemand(m_adj, use_domestic=TRUE)
	v$Failure
	# rownames variable
	# 399 S00402/US V1
	# 400 S00300/US V1
	# Test if model domestic output can be recalculated
	v <- compareOutputandLeontiefXDemand(m, use_domestic=TRUE)
	v$Failure
	# rownames variable
	# 399 S00402/US V1
	# 400 S00300/US V1
	# Test if adjusted model E can be recalculated
	v <- compareEandLCIResult(m_adj,use_domestic=FALSE,tolerance=0.02)
	v$Failure
	#0
	v$N_Pass
	#[1] 7218
	v <- compareEandLCIResult(m_adj,use_domestic=TRUE,tolerance=0.02)
	v$Failure
	#0
	v$N_Pass
	#[1] 7218
	# Compare result to standard model if E can be recalculated
	v <- compareEandLCIResult(m,use_domestic=FALSE,tolerance=0.02)
	v$Failure
	#0
	v$N_Pass
	#[1] 7236
	v <- compareEandLCIResult(m,use_domestic=TRUE,tolerance=0.02)
	v$Failure
	#0
	v$N_Pass
	#[1] 7236

[MoLi7]

Re: how Used goods are Scrap are handled in non-US IO tables

1. China: "Scrap and waste recycling products" encompassing activities involving the recovery of various waste streams—solid waste, wastewater (liquid), and waste gases—to make them suitable for conversion into new raw materials (source in Chinese). This includes:
   • Recycling and sorting of non‑metallic waste and scrap to prepare them for further processing into new raw materials.
   • Recovery of metal waste and scrap suitable for further processing into metal raw materials.
   • Dismantling and recycling of waste electrical and electronic equipment (WEEE/E‑waste).
2. UK: "Waste collection, treatment and disposal services; materials recovery services"
3. Japan: "Reuse and recycling"
4. Korea: "Resource Recycling Services"
5. Tier 2 countries: "Water supply; sewerage, waste management and remediation activities"

In summary, China, Japan, and Korea IOTs have relatively explicit sectors for Scrap (and recycled products), whereas the UK and all Tier 2 country IOTs aggregate Scrap with waste management even water supply.