1. 1 Econ 240C Power 17

2. 2 Outline The Law of One Price Moving averages as a smoothing technique: Power 11_06: slides 86-110 Intervention models: Power 13_06: slides 51-89

3. 3 Law of One Price: Outline Definition: slides 4-5 Applied to wheat trade: slides: slides 7-8 Time Series Notation: slides 9-10 Data and Traces: slide 11-15 Show that logs of import & export prices are evolutionary of order one: slides 16-25 Show that log of price ratio is stationary: slides 26-38 –Speed of convergence: slides 39-43

4. 4 Outline Cont. Cointegration: slides 44- –Long run equilibrium relationship between log of import price and log of export price (with freight and lags): slides 44-46 –VAR speed of adjustment model: slide 47

5. 5 The Law of One Price The New Palgrave Dictionary of Money and Finance –Next slide

6. 6 The Law of One Price This law is an immediate consequence of the absence of arbitrage and, like the absence of arbitrage, follows from individual rationality. Departures from the no arbitrage condition imply that there are profit opportunities. These arise because it would be profitable for arbitrageurs to buy good i in the country in which it is cheaper and transport it to the country in which it is more expensive, and in doing so, profit in trade.

7. 7 Commodity Trade Issues Well defined product: World Wheat Statistics –# 2 Dark Northern Spring 14% –Western White –Hard Winter Transport costs –US: export Pacific Ports Gulf Ports

8. 8 Prices in $/metric ton Import price notation: DNSJ is Japanese import price in $/metric ton for Dark Northern Spring wheat Export price notation: DNSG is export price for Dark Northern Spring from a Gulf Port; DNSP is export price for Dark Northern Spring from a Pacific Port –Lagged one month because commodity arbitrage takes time

9. 9 Transport Cost in $/Metric Ton Freight rates are forward prices and are lagged two months

10. 10 Time Series Import Price: DNSJ Export Price (lagged one) Plus Freight (lagged two): DNSGT Logarithm of Price Ratio: ln [DNSJ/DNSGT] = lnDNSJ – lnDNSGT denoted lnratiodnsgjt = ln[1 + ∆/DNSGT] ~ ∆/DNSGT, the fractional price differential, where ∆ = DNSJ – DNSGT, and can be positive or negative

11. 11 Time Series Is the log of the export price evolutionary, of order one? –Ln DNSJ = lndnsj Is the log of the import price evolutionary, of order one? –Ln DNSGT = lndnsgt Is their difference stationary, of order zero, ie. are they cointegrated? i.e. Is the log price ratio ( the fractional price differential) stationary? –Ln{DNSJ/DNSGT] =lnratiodnsjgt

12. 12 Data

13. 13 Table Three: Pacific DNS Log of Ratio of Import Price to Pacific Export Price (lag 1) Plus Pacific Freight (lag2) –Stationary No unit root AR (1) Model: root 0.54, normal residual Log of Import Price and Log of Pacific Export Price (lag 1) Plus Pacific Freight (lag 2) –Cointegrated VEC: one lag Rank: 1, 1, 1, 1, 2 Data: no trend; Integrating Equation: Intercept, no trend, rank one, 1%

14. 14

15. 15

16. 16

17. 17 Show that Logs of Prices Are Evolutionary

18. 18

19. 19

20. 20

21. 21 Conclude Log of Import Price, lndnsj, is evolutionary

22. 22

23. 23

24. 24

25. 25 Conclude Both the log of the import price and the log of the Pacific export price (lagged one) plus the Pacific Freight Rate (lagged two) are evolutionary, of order one. –To be of order one, not higher, their differences should be stationary, i.e. of order zero. –Unit root tests show this is the case

26. 26

27. 27 Lof of Price Ratio

28. 28

29. 29

30. 30

31. 31

32. 32

33. 33

34. 34

35. 35

36. 36

37. 37

38. 38 Conclusions Log of ratio of import price to the export price (lagged one) plus freight rate (lagged two) is stationary and is modeled as an autoregressive process of the first order with mean zero and root 0.54

39. 39

40. 40 How fast does any price differential get arbitraged to zero? Arone(t) = b*arone(t-1) + wn(t) Arone(t-1) = b* arone(t-2) + wn(t-1) Arone(t) = b[b* arone(t-2) + wn(t-1)] + wn(t) Arone(t) = b 2 *arone(t-2) + wn(t) + b*wn(t-1) Arone(t+2) = b 2 *arone(t) + wn(t+2) + b*wn(t+1) Arone(t+u) = b u *arone(t) + wn(t+u) + b*wn(t+u- 1) + …

41. 41 Half Life 1.1 month ~ 5 weeks Arone(t+u) = b u *arone(t) + wn(t+u) + b*wn(t+u-1) + …. E t {Arone(t+u) = b u *arone(t) + wn(t+u) + b*wn(t+u-1) + …} E t Arone(t+u) = b u *arone(t) E t Arone(t+u) /arone(t) = ½ = b u Ln [Arone(t+u) /arone(t)] = ln(1/2) = u*lnb - 0.693/lnb= -0.693/ln 0.54 = 1.1 = u

42. 42

43. 43 Half Life: root =0.54 Time =uLog ratiobubu 01001 1540.54 229.160.2916 315.74640.157464

44. 44 Cointegration Logs of export price and import price (lagged with freight lagged) are of order one. Their difference is of order zero The long run relationship: –Lndnsj = c + b*lndnspjt + e –Where the residual is an estimate of price differential over time

45. 45

46. 46

47. 47 Error Correction VAR dlndnsj(t)= a M *e(t-1) + wn M (t) + b 11 dlndnsj(t-1) + c 12 dlndnspjt(t-1) dlndnspjt(t)= -a x *e(t-1) + wn x (t) + b 21 dlndnsj(t-1) + c 22 dlndnspjt(t-1) a M and a x are speed of adjustment parameters of fractional change in import and export prices to the fractional price differential, i.e. e(t-1)

48. 48

49. 49 Significant speed of adjustment Parameter. If lndnsj is greater than the fitted value, i.e. Greater than c + b*lndnspjt, the Residual e(t) is positive, and next Period, lndnspjt will increase to Close the gap.

50. 50 Johansen Cointegration Test

51. 51 Johansen Table Summary

52. 52 Diebold, Ch.4 p. 87

53. 53

54. 54

55. 55 Null: No Cointegrating Equation Reject null at 1% level

56. 56 Impulse response functions

57. 57 Impulse response functions

58. 58 Variance decomposition

59. 59 Error Correction Model: All the way in one play

60. 60

61. 61 Error Correction VAR, One Lag

62. 62 VEC Cont.

63. 63 Johansen Summary Table

64. 64 Johansen Test: No Data Trend, Intercept

65. 65 Impulse Response Functions Order: lndnsj, lndnspjt

66. 66 Impulse Response Functions Order: lndnspjt, lndnsj

67. 67 Variance Decomposition Order: lndnsj, lndnspjt

68. 68 Table 3: log ratio of Import Price/ [Export Price (lag 1) + Freight (lag 2)] DNSGJTDNSPJTWWPJTHWGJT AR(1)0.630.540.490.39 MA(1)0000 Model Resnormal VEC1 lag 5 Mod. rank0,0,1,0,21,1,1,1,2 1,1,1,1,1 Rank 15%1% (3/4) 1% all

69. 69 Table 4: log ratio of Import Price/ [Export Price (lag 1)] DNSGJDNSPJWWPJHWGJ AR(1)0.740.710.730.64 MA(1)0000 Model Resnormal VEC1 lag 5 Mod. rank0,0,0,0,20,0,0,0,0 0,0,1,0,2 Rank 15%