1. the Global Value-Chain Training and Research Workshop
Computable general equilibrium (CGE) modelling and global supply chain (GSC) analysis
presentation by
Peter B. Dixon at
the Global Value-Chain Training and Research Workshop
UIBE , Beijing
August 2, 2017

2. Topics Input-output economics, the foundation of CGE & GSC
(a) the input-output table
(b) national accounts
(c) the one-country input-output model
(d) the Leontief paradox
(e) effective rates of protection
(e) the multi-regional or multi-national I-O model
- starting point for CSC analysis
2. CGE modelling
(a) Johansen’s model of Norway
(b) Johansen’s solution matrix versus Leontief’s inverse
(c) trade in a CGE model, Armington to Melitz

3. Topics 3. Applying existing CGE models to supply chain issues
(a) Walmsley & Minor on reversing NAFTA
(b) Dixon, Rimmer and Waschik on Buy-America(n)
Towards a CGE model with supply chain features: designs,
components, assembly and sales
5. Concluding remarks

4. 1. Input-output economics
Wassily Leontief
Nobel prize for economics
in 1973

5. Input-output database for USAGE, a detailed CGE model of the US

6. National accounts: value of inputs equals value of output
Simplifying:

7. Income and expenditure measures of GDP

8. One-country input-output model

9. Leontief paradox
for the US in 1947

10. Corden’s effective rates of protection

11. Multi-regional input-output model

12. Calculating how much value added from each country is embodied in s’s exports
The first term on the RHS is s’s own contribution to value added in its exports (known as domestic value added, DVs). The second term is foreign value added (FVs).

13. How much value added does country s really export?
VA exported from s, and returned in final goods
VA exported from s, and returned in interm goods
Exports from s, VA content
VA from s, absorbed outside s
Foreign VA in exports from s = + + +
Koopman, R., Z. Wang and S-J. Wei (2014), “Tracing value added and double counting in gross exports”, American Economic Review, 104(2),

14. Relating fob exports to VA content of exports: tariff correction

15. 2. CGE modelling
What makes a CGE model: defining features
Computable: real world data
General: explicit agents with optimizing behaviour
Equilibrium: price determination through the interaction of demand and supply for commodities and factors (introduces resource constraints)
Popular - 15,000 practitioners from 150 countries in the GTAP network
Versatile - the effects on
macro, welfare, industry, regional, labour-market, distributional and environmental variables of
taxes, public consumption and social security payments; tariffs and other interferences in international trade; environmental policies; technology; international commodity prices; interest rates; wage setting arrangements and union behaviour; mineral discovories (the Dutch disease); immigration; micro- economic reform; and major projects

16. The first CGE model: Johansen for Norway
Leif Johansen

17. Johansen put equations in linear percentage change form

18. Johansen’s T matrix 46 by 86

19. Johansen’s CGE model versus Leontief’s I-O model:
Effects of demand stimulation
Leontief: X=(I-A)-1*y Johansen: v1 = T*v2
all  0
mainly > 0
all  0
mainly > 0
complementary
all 1
mainly < 0
mainly < 0
competition
all 0 but mainly <1
T(x,z) is the 22 by 22 sub-matrix of T showing elasticities of outputs with respect to exogenous demands

20. Trade in CGE models
Armington’s 1969 specification for MERM introduced to CGE modelling via the ORANI model of Australia and the Michigan multi-country model
destination
Com
Source

21. Modern trade theory in CGE modelling: Melitz
Firm heterogeneity
Monopolistic competition
Economies of scale
Love of variety
Melitz welfare effects of tariff change
= Armington effects
+ extra effects from changes in variety
and in intra-industry efficiency
Marc Melitz
Melitz specifications are not leading to conclusions that are noticeably different from Armington: extra effects cancel out reflecting Melitz optimality

22. 3. Applying existing CGE models to supply chain issues
For analysis of policy questions associated with supply chains we want a model that identifies:
users (u) in destination countries (d) of finely disaggregated intermediate inputs supplied from different source countries (s)
GTAP: many source countries (s) and destination country (d) but only aggregated products (e.g. Motor vehicles & parts, Mvh) and no identification of users (u)
Walmsley and Minor have recently added the u dimension:
Now they can answer questions of the form: does an increase in U.S. tariffs on Mvh hurt or help the U.S. Mvh industries?

23. USAGE analysis of Buy America: a supply chain issue
USAGE is a 400-sector CGE model of the U.S.: strong on commodity and user detail, but covers only one country.
Buy America
U.S. governments purchase mining, construction and manufacturing goods worth about $800 billion per year.
Buy America is intended to guide suppliers to the public sector of mining, construction and manufacturing goods towards domestically produced inputs. The aim is to protect U.S. industries, particularly manufacturing industries such as steel, against import competition.
Buy America interferes with efficient supply chain arrangements especially between the U.S. and Canada. Trade Partnership Worldwide (TPW, 2016) gives examples for North American firms manufacturing cars, buses, steel girders, fabricated metal products and plumbing materials. In each case, firms have integrated supply chains spreading across the U.S. and Canada.

24. USAGE analysis of Buy America
We assume that Buy America is comprehensively in place and simulate the effects of removing it.
In formulating the impact effects of removing Buy America (the shocks) we examine how the import-domestic mix of each intermediate input flow would be affected.
Dixon, P.B., M.T. Rimmer and R.G. Waschick (2017), “Evaluating the effects of local content measures in a CGE model: eliminating the U.S. Buy America(n) programs”, Economic Modelling, forthcoming.

25. Sales of Plumbing materials (C78) into Power & communication structures (I29), $million and percentages

26. Sales of goods to U.S. goods-producing industries, $ million and percentages (aggregation over 259 by 389 input-output cells)

27. Micro effects of scrapping Buy America
Moves import shares for government production towards those for private
production
Gives efficiency gain for government production

28. Original situation Sales of Plumbing materials (C78) into Power & communication structures (I29), $million and percentages

29. After scrapping Buy America(n) Sales of Plumbing materials (C78) into Power & communication structures (I29) freed from Buy America(n), $million and percentages

30. Simulating the scrapping of Buy America
Shocks: technical changes that favor imported inputs relative to
domestic inputs but generate efficiency gains
Macro assumptions:
(a) real wages fixed, aggregate capital and investment fixed
How many more jobs could the U.S. economy support with
its current wages and resources?
(b) public consumption fixed, indirect taxes reduced to keep
public-sector deficit fixed
(c) balance of trade fixed via flexible exchange rate
(d) terms of trade fixed - policy is implicitly part of a free-trade
agreement

31. Macro effects of scrapping Buy America, %

32. Commodity output effects (%) of Buy America(n) cessation: USAGE and Fitted results and explanatory variables

33. Commodity output effects (%) of scrapping Buy America(n): USAGE & fitted results

34. Employment effects by industry of Buy America cessation: changes and percentage changes in jobs

35. Employment effects by industry of Buy America cessation: changes and percentage changes in jobs

36. 50 winning states out of winning congressional districts out of Employment effects by state of scrapping Buy America programs

37. Buy America conclusions

38. Towards a supply chain CGE model: Design, Components, Assembly and SalesDist for widgets

39. Solving the widget model
27 global production scenarios (s=1, 2,… 27):
Consider each scenario. Assume that the 2-country economy will choose the one that gives minimum costs of satisfying given global demands (X41 and X42) for widgets. Thus to solve the model we must find s to
minimize Cost(s)
where

40. To evaluate Cost(s) we must determine P41(s) and P42(s) from the equation:

41. Evaluating Cost(s): step 1, determining commodity outputs by country

42. Evaluating Cost(s): step 2, determining scaled unit labor costs

43. Evaluating Cost(s): step 3, determining input-output coefficients, Ard(s)

44. Evaluating Cost(s): step 4, use p = pA+v to determine prices including P41(s) and P42(s)
giving
Now we can evaluate Cost(s) = P41(s)*X41 + P42(s)*X42 for each s to determine the cost-minimizing scenario (or possibly scenarios). This gives us the solution (or possibly solutions) for outputs, employment and trade.

45. Pre GSC (e.g. 1990)
Assumptions
Results

46. Pre GSC (e.g. 1990): further results
Koopman, Wang & Wei decomposition of exports
Pre GSC, only country 1 undertakes design, components, assembly.
Country 2’s only activity in the widget industry is internal sales and distribution.
Trade is one-way: fully assembled products are exported from country 1 to country 2. We can refer to this as traditional or Ricardian trade.

47. GSC conditions start to emerge (e. g
GSC conditions start to emerge (e.g. 1995) lower T’s and UC’s in country 2
Assumptions
Results

48. GSC conditions start to emerge (e. g
GSC conditions start to emerge (e.g. 1995) lower T’s and UC’s in country 2
Further results:
Between 1990 and 1995, unit costs in country 2 fall sharply and trade costs in both directions also fall. Country 2’s unit costs for components (0.9) are now lower than in country 1 (1.0). Yet country 1 continues to be the only country that produces design, components & assembly.
Why doesn’t country 2 take over production of components?
Imagine that components are produced in country 2, but assembly and design remain in country 1.
The saving of component VA is 10% of the initial world value added in components, a saving of
However, we now have to ship 1.75 of design with value 1.75 from 1 to 2. This incurs a trade cost of (=1.75*0.05). Also we have to ship 1.75 of components with value (= [0.95* *0.95]*1.75) from 2 to 1. This incurs a trade cost of (=3.242*0.05). In total, the extra trade costs from relocating components alone to country 2 are 0.250, greater than the potential saving of

49. MTR diagram for analysing Material Trade Routes. .
Country 1
Country 2
SalesDist
1.00
0.75
0.75 . .
1.00
1.00
0.75
1.75
Assembly . .
1.75
1.75
1.75
Components . .
1.75
1.75
Design
1.75
Why doesn’t country 2 produce all the components when its UC for components falls to 0.9?
To work out the answer, compare production and trade costs on the red paths and the dotted black paths.

50. GSC emerges (e.g. 2000) lower UC for assembly in country 2
Assumptions
Results

51. GSC emerges (e.g. 2000) lower UC for assembly in country 2
Further results:

52. MTR diagram for analysing Material Trade Routes. .
Country 1
Country 2
SalesDist
1.00
0.75 . .
1.00
1.00
0.75
0.75
1.00
0.75
Assembly
1.75
1.00 . .
0.75
1.75
1.75
Components
1.75 . .
1.75
Design
1.75
Why does the reduction in UC assembly costs in country 2 from 1.25 to 1.15 trigger the establishment of world production of components in 2 and self-sufficient production of assembly?
To work out the answer, compare production and trade costs on the red paths and the dotted black paths.

53. GSC fully developed (e. g
GSC fully developed (e.g. 2010) assembly UC in 2 falls well below that for 1
Assumptions
Results

54. GSC fully developed (e. g
GSC fully developed (e.g. 2010) assembly UC in 2 falls well below that for 1
Further results:

55. MTR diagram for analysing Material Trade Routes. .
Country 1
Country 2
SalesDist
1.00
0.75 .
1.00 .
1.00
0.75
0.75
0.75
1.00
Assembly
1.75
1.00 .
1.75 .
0.75
1.75
Components
1.75 . .
1.75
Design
1.75
Why does there need to be a large reduction in UC assembly costs in country 2 (from 1.15 to 0.90) before country 2 becomes dominant in assembly?
To work out the answer, compare production and trade costs on the blue paths and the dotted black paths.

56. Summary of results from the supply chain widget model

57. Pioneer of inframarginal economics
5. Concluding remarks
Leontief’s I-O table is the outstanding contribution to applied economics of the 20th century
Starting with Johansen, CGE has become a key tool for policy analysis
Quantitative GSC analysis has been largely descriptive. The next step is to build models for elucidating the effects of policy and other changes in trading arrangements in a GSC world
CGE has been applied to GSC issues, but captures only weakly elements of global supply chains
A satisfactory CGE model incorporating GSC will describe the footloose nature of design, component and assembly activity. This will be necessary for providing worthwhile advice on the distribution of jobs across activities & countries in industries producing traded goods
The widget GSC model suggests results outside the range of conventional CGE. But the widget results are too extreme. As Dixon (2006) showed in an analysis Xiaokai Yang’s inframarginal economics, aggregation over heterogeneous firms or households can smooth out at the industry level extreme movements in variables for individual agents
Xiaokai Yang
Pioneer of inframarginal economics
Dixon, P.B. (2006), “Inframarginal economics: an outsider’s view”, Economic Papers, Vol. 25(2), June, pp