WEBVTT

NOTE Inventories, Buffer Stocks, and Excess Capacity: Fractal Analysis of Antipersistence, Entrepreneurial Management, and Austrian Business Cycle Theory

1
00:00:00.000 --> 00:00:13.000
What does Austrian business cycle theory imply for buffer stocks, the necessary stockpiles of raw materials, goods in process and finished intermediate output required for each stage of the production process?

2
00:00:13.000 --> 00:00:21.000
The lower interest rates brought about by credit expansion force the accumulation of buffer stocks in earlier stages of production.

3
00:00:21.000 --> 00:00:26.040
Because lower interest rates also result in a reduction in saving and a

4
00:00:26.040 --> 00:00:30.640
corresponding increase in consumer spending, buffer stocks also accumulate in

5
00:00:30.640 --> 00:00:35.640
late stages of production. So buffer stocks are progressively reallocated to

6
00:00:35.640 --> 00:00:40.640
early and late stages being transferred there from middle stages. This

7
00:00:40.640 --> 00:00:46.480
unsustainable expansion leads to recession. How can we measure buffer

8
00:00:46.480 --> 00:00:50.960
stocks? In this paper I use the ratio of specific sectoral producer prices

9
00:00:50.960 --> 00:00:54.700
Price Indices divided by the Consumer Price Index.

10
00:00:54.700 --> 00:01:00.400
This ratio increases when the demand for sectoral buffer stocks rises faster than for consumption

11
00:01:00.400 --> 00:01:06.860
and falls when demand for consumption rises faster than for sectoral buffer stocks.

12
00:01:06.860 --> 00:01:09.680
There are at least two alternative measures.

13
00:01:09.680 --> 00:01:14.360
Buffer stocks can be measured directly as inventories of goods in process or intermediate

14
00:01:14.360 --> 00:01:20.600
output for each sector, or we could use capacity utilization rates for each sector.

15
00:01:20.600 --> 00:01:26.380
Neither was used in this paper, though either might be used in subsequent research.

16
00:01:26.380 --> 00:01:30.000
Regressions were estimated where the buffer stock measure was explained by the three-month

17
00:01:30.000 --> 00:01:32.020
Treasury bill rate.

18
00:01:32.020 --> 00:01:35.940
Since credit expansion lowers the interest rate, the slope on interest should be negative

19
00:01:35.940 --> 00:01:40.860
for early and late stages of production and positive for middle stages.

20
00:01:40.860 --> 00:01:45.180
These regressions, unfortunately, had very low r-squares.

21
00:01:45.180 --> 00:01:50.460
Higher r-squares were obtained by adding 10-year Treasury bill rates, significant negative

22
00:01:50.460 --> 00:01:55.540
Positive coefficients were found on 10 out of 18 sectors suggesting these sectors are

23
00:01:55.540 --> 00:02:00.900
either early or late based on the result that the buffer stock measure rises when interest

24
00:02:00.900 --> 00:02:05.260
rates fall, that is in response to credit expansion.

25
00:02:05.260 --> 00:02:09.500
This doesn't offer a statistical basis to distinguish early from late-state sectors

26
00:02:09.500 --> 00:02:11.340
however.

27
00:02:11.340 --> 00:02:15.780
Significant positive coefficients were found on 7 out of 18 sectors suggesting these are

28
00:02:15.780 --> 00:02:18.280
Intermediate Stages of Production.

29
00:02:18.280 --> 00:02:20.500
And that leaves out one which didn't have

30
00:02:20.500 --> 00:02:22.900
either coefficient significant.

31
00:02:23.820 --> 00:02:26.500
None of these results is too surprising.

32
00:02:26.500 --> 00:02:30.220
The question marks are indicating ones

33
00:02:30.220 --> 00:02:32.920
that I thought weren't completely obvious

34
00:02:32.920 --> 00:02:35.000
that they should necessarily be late,

35
00:02:35.000 --> 00:02:38.020
early stage sectors, but I didn't think

36
00:02:38.020 --> 00:02:39.780
it was such a bad outcome.

37
00:02:40.660 --> 00:02:43.460
More question marks here for the intermediate stages.

38
00:02:43.460 --> 00:02:45.460
Why some stages that we would expect

39
00:02:45.460 --> 00:02:50.060
A priori to be late stages appear here is problematic.

40
00:02:50.060 --> 00:02:53.340
For the late stages, which again, I arbitrarily decided

41
00:02:53.340 --> 00:02:56.620
this is a late stage versus this is an early stage,

42
00:02:56.620 --> 00:02:59.900
based on those that I got the outcome,

43
00:02:59.900 --> 00:03:03.180
these should be 10 out of 18 are early or late.

44
00:03:04.020 --> 00:03:06.860
All right, and there are no surprises here.

45
00:03:06.860 --> 00:03:09.140
Finally, the residuals from these regressions

46
00:03:09.140 --> 00:03:12.460
are subjected to power spectral density analysis.

47
00:03:12.460 --> 00:03:14.660
These residual series should be stationary

48
00:03:14.660 --> 00:03:18.560
and our linear functions of three stationary series.

49
00:03:18.560 --> 00:03:21.900
The buffer stock measure which is a ratio

50
00:03:21.900 --> 00:03:23.760
of the producer price index for the sector

51
00:03:23.760 --> 00:03:26.220
divided by the consumer price index

52
00:03:26.220 --> 00:03:28.220
and the two interest rates.

53
00:03:28.220 --> 00:03:30.560
The finding that the residuals are antipersistent

54
00:03:30.560 --> 00:03:32.700
with Hearst exponent less than one half

55
00:03:32.700 --> 00:03:35.200
indicates that the sector is dominated

56
00:03:35.200 --> 00:03:38.380
by interest rate movements which we would expect

57
00:03:38.380 --> 00:03:43.380
for normal, I'm sorry, for early sectors.

58
00:03:44.660 --> 00:03:57.660
The finding that residuals are persistent with first exponent greater than one half suggests the sector is dominated by normal ongoing entrepreneurial adjustment, which should be characteristic of late sectors.

59
00:03:57.660 --> 00:04:08.660
And if it's very close to one half, that indicates that the sector is dominated by normally distributed adjustments for middle stages.

60
00:04:08.660 --> 00:04:21.000
To perform spectral density analysis, the series is divided into various frequencies or subsample sizes analogous to interval lengths or windows in the rescale range and roughness length methods.

61
00:04:21.000 --> 00:04:29.200
The signal power, a Fourier transform measure of variation occurring within each wavelength, is computed for each frequency.

62
00:04:29.200 --> 00:04:40.200
In the log-log plot of this power measure graphed against the frequency, the slope is estimated by least squares as negative b, which is then used to compute h, the Hurst exponent.

63
00:04:40.200 --> 00:04:52.200
These are two examples. This compares the lowest sloped power spectrum graph for crude materials on the left with the highest for finished goods, less food and energy on the right.

64
00:04:52.200 --> 00:05:00.200
and because it has a low slope, it's significantly less than one half, we can infer anti-persistence

65
00:05:00.200 --> 00:05:05.480
for the adjustments in the crude materials buffer stock, which is and would be expected

66
00:05:05.480 --> 00:05:13.080
a priori to be an early stage sector of production and for finished goods, less food and energy

67
00:05:13.080 --> 00:05:18.760
and pretty much anything that's described as finished goods, finished consumer goods,

68
00:05:18.760 --> 00:05:23.480
However categorized, you would expect that's fairly likely to be a late stage of production,

69
00:05:23.480 --> 00:05:28.240
although that outcome is not absolutely universal.

70
00:05:28.240 --> 00:05:33.280
So based on the spectral density analysis, we see clearly that the earliest stages have

71
00:05:33.280 --> 00:05:38.120
the lowest Hearst exponents, meaning that the buffer stocks are anti-persistent and

72
00:05:38.120 --> 00:05:40.480
driven by credit expansion.

73
00:05:40.480 --> 00:05:45.200
Why finished consumer foods appears here is a bit of a mystery.

74
00:05:45.200 --> 00:05:51.740
The latest stages have the highest Hurst exponents and these results are intuitively appealing.

75
00:05:51.740 --> 00:05:57.760
Middle stages are defined somewhat arbitrarily as those with Hurst exponents closest to one

76
00:05:57.760 --> 00:06:03.420
half and the arbitrary limit that I set was within one sample standard deviation of one

77
00:06:03.420 --> 00:06:09.200
half. However, this appears to include some late stages which is a bit of a puzzle. This

78
00:06:09.200 --> 00:06:15.180
may occur because the producer price indices track the consumer price index fairly closely

79
00:06:15.180 --> 00:06:21.180
Mostly for Late Stages, which then would impose normality, the result that H is equal to one half.

80
00:06:21.180 --> 00:06:28.180
It's more difficult to explain why this category includes some apparently late stages.

81
00:06:28.180 --> 00:06:33.180
These results overall support and are mostly very easy to interpret according to Austrian Business Cycle Theory,

82
00:06:33.180 --> 00:06:36.180
although there are a few idiosyncratic quirks.

83
00:06:36.180 --> 00:06:42.180
However, the finding of anti-persistence in 10 out of 18 sectors is very, very difficult to reconcile

84
00:06:42.180 --> 00:06:45.180
with the alternative real business cycle theory.

85
00:06:45.180 --> 00:06:49.180
Does that music mean that my time is up?

86
00:06:49.180 --> 00:06:52.180
And I have to get off the stage.

87
00:06:52.180 --> 00:06:54.180
Thank you very much.