1: \documentclass[12pt]{article}
2: \usepackage{psfig,times,fancyheadings}
3:
4: %%%%% Margins for American paper format (JeanL)
5: % \topmargin= 0.2cm \oddsidemargin = +0.4cm \evensidemargin = +0.4cm
6: %%%%% Margins for European paper format (JeanL)
7: \topmargin=-0.7cm \oddsidemargin = -0.4cm \evensidemargin = -0.4cm
8:
9: %\textheight=237mm \textwidth=170mm \baselineskip 5mm
10: \textheight=247mm \textwidth=170mm \baselineskip 5mm
11: \pagestyle{fancyplain}
12:
13: \parindent=0pt
14:
15: \begin{document}
16: \thispagestyle{empty}
17:
18: % \lhead[\fancyplain{}{\sl Price waves}]{\fancyplain{}{\sl The
19: % Newsletter of Econophysics}}
20: % \rhead[\fancyplain{}{\sl The Newsletter of Econophysics}]{\fancyplain{}{\sl
21: % April 1997 Number 1}}
22:
23: \lhead[\fancyplain{}{\sl }]{\fancyplain{}{\sl }}
24: \rhead[\fancyplain{}{\sl }]{\fancyplain{}{\sl }}
25:
26: %%%%% Definitions
27:
28: \newcommand{\nc}{\newcommand}
29:
30: \nc{\qI}[1]{\section{ {#1} }}
31: \nc{\qA}[1]{\subsection{ {#1} }}
32: \nc{\qun}[1]{\subsubsection{ {#1} }}
33: \nc{\qa}[1]{\paragraph{ {#1} }}
34:
35:
36:
37: \nc{\qfoot}[1]{\footnote{ {#1} }}
38: \def\qL{\hfill \break}
39: \def\qpar{\vskip 2mm plus 0.2mm minus 0.2mm}
40: \def\tvi{\vrule height 12pt depth 5pt width 0pt}
41:
42: \def\qparr{ \vskip 1.0mm plus 0.2mm minus 0.2mm \hangindent=10mm
43: \hangafter=1}
44:
45: % Decale UN paragraphe
46: % Attention! La double accolade est vitale, sinon tout le
47: % est decale (cf TEX p.199)
48: % On peut aller a la ligne avec \qL=\hfill \break
49: % Par contre ne supporte pas les lignes blanches
50: \def\qdec#1{\par {\leftskip=2cm {#1} \par}}
51:
52:
53: %% Defs specifiques
54: \def\qdpt{\partial_t}
55: \def\qdpx{\partial_x}
56: \def\qddpt{\partial^{2}_{t^2}}
57: \def\qddpx{\partial^{2}_{x^2}}
58: %
59: \def\qn#1{\eqno \hbox{(#1)}}
60: \def\qds{\displaystyle}
61: \def\qal{\sqrt{1+\alpha ^2}}
62: \def\qw{\widetilde}
63:
64:
65:
66: %%%%% End of definitions
67:
68: \null
69: \vskip 1cm
70: {\bf \sl \large To appear in: The European Physical Journal B}
71: \vskip 2cm
72:
73: \centerline{\bf \LARGE Determining bottom price-levels}
74: \vskip 0.5cm
75: \centerline{\bf \LARGE after a speculative peak}
76:
77: \vskip 1cm
78: \centerline{\bf B.M. Roehner $ ^* $ }
79: \centerline{\bf L.P.T.H.E. \quad University Paris 7 }
80:
81: \vskip 2cm
82:
83: {\bf Abstract}\quad During a stock market peak the price of a given
84: stock ($ i $) jumps from an initial level $ p_1(i) $ to a peak level
85: $ p_2(i) $ before falling back to a bottom level $ p_3(i) $. The
86: ratios $ A(i) = p_2(i)/p_1(i) $ and $ B(i)= p_3(i)/p_1(i) $ are referred to
87: as the peak- and bottom-amplitude respectively. The paper
88: shows that for a sample of stocks there
89: is a linear relationship between $ A(i) $ and $ B(i) $ of the
90: form: $ B=0.4A+b $. In words, this means that the higher the price of
91: a stock climbs during a bull market the better it resists during the
92: subsequent bear market. That rule, which we call the resilience pattern,
93: also applies to other speculative markets.
94: It provides a useful guiding line for Monte Carlo simulations.
95:
96:
97:
98: \vskip 1cm
99: \centerline{\bf 5 June 2000 }
100: % \centerline{\bf 2 May 1999 \quad First draft \quad Comments welcome}
101:
102: \vskip 1cm
103: {\bf PACS.}\ 64.60 Equilibrium properties near critical points -
104: 87.23Ge Dynamics of social systems
105:
106:
107: \vskip 2cm
108: * Postal address: LPTHE, University Paris 7, 2 place Jussieu, 75005 Paris, France
109: \qL
110: \phantom{* }E-mail: ROEHNER@LPTHE.JUSSIEU.FR
111: \qL
112: \phantom{* }FAX: 33 1 44 27 79 90
113:
114: \vfill \eject
115:
116:
117:
118: \qI{Introduction}
119:
120: Traffic jams are fairly unpredictable because they depend upon a large
121: number of factors, e.g. timing in the traffic, weather conditions,
122: highway maintenance, automobile accidents, some of which are completely
123: random. However once begun, traffic jams display fairly recurrent patterns
124: as to average duration, behavior of the drivers, and so on. This
125: traffic jam parallel has been introduced by Charles Tilly [12] in
126: the context of historical sociology in order to explain why revolutions
127: can hardly be predicted. It also applies to the occurrence of speculative
128: price peaks: the downturn of price peaks can hardly
129: be predicted because they depend upon a number of (possibly)
130: exogenous factors%
131: \qfoot{For instance it has been argued (Business Line 8 May 2000) that
132: the spurt in the NASDAQ composite index that occurred in December 1999 and
133: January 2000 was fueled by a Y2K-motivated injection of money into the
134: banking system. Needless to say, no model will ever be able to take
135: into account such circumstantial factors. However, the present paper
136: suggests that when a market rallies or plummets
137: there are some structural invariants
138: which are independent of triggering factors.}%
139: .
140: However, as is the case for traffic jams, once a price peak is under
141: way it obeys some definite rules; accordingly its outcome can to some
142: extent be predicted at the level of individual companies.
143: \qL
144: More precisely
145: we focus our attention on the relationship between the prices at the
146: beginning of the peak, at the peak and at the end of the peak. We
147: denote by $ p_1 $ the price level at the start of the peak, by $ p_2 $
148: the price at the peak and by $ p_3 $ the bottom price at the end
149: of the falling price path; we further introduce the peak amplitude
150: $ A = p_2/p_1 $ and the bottom amplitude $ B = p_3/p_1 $.
151: $ A $ and $ B $ can be defined for any company in the market; for instance
152: on the NASDAQ where there are currently more than 5,000 companies listed
153: $ A $ and $ B $ can be seen as variables for which there
154: are several thousand realizations. In the next section we show that
155: $ A $ and $ B $ are closely correlated; with a correlation of the order
156: of 0.75 the following regression holds:
157: $$ B = aA + b \qn{1} $$
158:
159: where $ a $ is usually of the order of 0.4. The fact that $ a $ is positive
160: means that the higher the peak amplitude, the larger the bottom amplitude;
161: in other words the higher the price of a stock climbs during the
162: rising phase (bull market) the better it resists during the falling
163: phase (bear market). The regularity summarized by equation (1) will
164: be referred to as the resilience pattern.
165: \qL
166: The paper proceeds as follows. In the second section the statistical
167: methodology is explained, then the resilience pattern is established
168: and illustrated through several case studies: we consider three
169: stock market peaks, one price bubble for postage stamps and two
170: real estate bubbles. In the conclusion we discuss the possible
171: implications of the resilience pattern.
172:
173: \qI{The resilience pattern}
174: \qA{Methodology}
175: Let us consider a typical price peak such as the one in Fig.1.
176: While it is easy to identify the moment $ t_2 $ when the price reaches
177: its peak level $ p_2 $, the determination of the initial moment $ t_1 $
178: (corresponding to $ p_1 $) and the end moment $ t_3 $ (corresponding
179: to $ p_3 $) is not so easy. Fortunately, as will be seen subsequently,
180: the relationship (1) does not depend upon the choice of $ t_1 $ or
181: $ t_2 $ in a critical way.
182: \qL
183: A peak will be delimited in two steps (i) At the global level of the
184: whole market the price peak is identified by using a broad annual
185: index; in that way one already gets a rough definition of
186: $ t_1, t_2, t_3 $ (ii) At the level of individual companies $ t_1 $
187: will be selected as the first year for which the
188: annual price change is positive,
189: $ t_2 $ as the peak year and $ t_3 $ as the last year for which the
190: annual price change is negative. Let us illustrate the procedure on an
191: example (Fig.1). As one knows there was a short bull market on the
192: NYSE after World War I which culminated in 1920;
193: thus, for the great majority of the stocks the first positive price change
194: was 1921-1922, which leads us to $ t_1=1921 $. At the end of 1929 the
195: downturn occurred very abruptly which means that for almost all stocks
196: $ t_2=1929 $. By and large the market bottomed out in 1931; yet for some
197: stocks such as for instance Consolidated Edison
198: the fall continued until 1935; in that case one would take $ t_3=1935 $.
199: \qL
200: Once the limits of the peak have been
201: determined for each individual stock,
202: the ratios $ p_2/p_1 $ and $ p_3/p_1 $ are computed for
203: the deflated prices; then the correlation and regressions are
204: carried out for the sample of stocks under consideration. We now
205: apply this procedure to several case studies.
206:
207: \qA{Stock markets}
208: As far as stock markets are concerned there is often a tendency to
209: over-emphasize the importance of crashes, by which we understand
210: a rapid fall occurring within one or two weeks, at the expense of the
211: long and steady declines that (in some cases) follow the crash. Crashes
212: are impressive because of their suddenness, but it is not obvious
213: and probably not even true, that crashes have a determining influence
214: on the medium-term (i.e. yearly) evolution of markets. A slide that
215: continues for over two or three years may have more significance for
216: the stock market and for the rest of the economy
217: than the crash itself.
218: A case in point is the parallel between the crashes of
219: October 1929 and October 1987 on the NYSE (Fig.2). The price paths were
220: very similar during the crashes and the six subsequent months;
221: but, as one knows, the ultimate outcomes were very different.
222: This observation
223: suggests that
224: steady slides (which are the topic of this paper) and abrupt crashes
225: are two different phenomena.
226: \qL
227: Applying the procedure delineated above, we obtain the results given
228: in Table 1. Note that the downturn in Paris occurred in February 1929
229: and cannot therefore be considered as a consequence of the Wall Street
230: crash; incidentally, in Germany the downturn took place in June 1928
231: that is to say more than one year before the downturn on the NYSE.
232: The analysis for the NYSE is based on the behavior of 85 individual
233: stocks. The correlation is 0.87 for
234: equation (1) and the distribution of sample points
235: in the $ (A,B) $ plane is shown in Fig.3; it displays the range
236: of both amplitudes and permits to verify that there is no non-linear
237: effect.
238: \qL
239: Since this is the largest sample considered in Table 1, it can
240: can be of interest to take a closer look at the statistical distribution
241: of the amplitudes: $ A $ and $ B $ have an average of 5.6 and 2.0
242: respectively and the standard deviations are 5.0 and 2.3. Moreover it
243: turns out that both $ A $ and $ B $ are distributed according to
244: a log-normal density. This could have been expected; indeed,
245: stock prices follow a log-normal law,
246: at least in first approximation
247: that is to say for time-samples of moderate size
248: and time intervals larger than one day,
249: and one knows that the ratio of two log-normal random variables
250: is also a log-normal random variable.
251: \qpar
252:
253: The analysis of the Paris stock market is not based on individual stocks
254: but on a set of indexes corresponding to 19 different economic sectors,
255: e.g. banks, coal mines, railroads, electricity, chemicals, etc. Some indexes
256: comprise more individual stocks than others, for instance the bank index
257: comprises 20 banks while the electricity index has only 11; on average there
258: are about 14 companies per sector. The results for the $ B $ versus
259: $ A $ regression are given in Table 1: the $ a $ estimate is fairly
260: close to the one obtained for the NYSE. The analysis of the 1989
261: peak on the Tokyo market is also based on indexes corresponding to
262: different economic sectors.
263: With as many as 26 different sectors the classification given in the
264: Japan Statistical Yearbook is even more detailed than the previous one.
265: It is not obvious whether that peak began in 1985 or in 1980.
266: It is true that the increase between 1980 and 1985 was not monotonic,
267: but one can argue with good reason that these fluctuations were
268: rather circumstantial. It is reassuring to observe that by taking
269: $ t_1=1980 $ one is lead to estimates which are fairly similar to those
270: obtained for $ t_1=1985 $; the fact that the correlation is higher
271: in the first case
272: in fact suggests that $ t_1=1980 $ is the most ``natural'' starting point.
273: \qpar
274:
275: On the basis of the resilience pattern it could seem that one can
276: invest in high-growth companies without much risk. This is not completely
277: true however for that rule only concerns the medium-term
278: behavior in the
279: vicinity of a given peak; it does not guarantee that in the long-run the
280: price of a stock which has experienced a huge peak will continue to increase.
281: A spectacular counter-example is shown in Fig.4. Not only did the
282: price of Columbia Gas System never again reach the level it had
283: attained in 1929 but it remained far below.
284:
285: \qA{Other speculative markets}
286: The consideration of other speculative markets relies on the implicit
287: assumption that the {\it basic} mechanisms of speculation are similar
288: for any speculative market. In this paragraph we examine the cases
289: of postage stamp and real
290: estate markets. These markets are particularly suited
291: for this kind of investigation because (i) they have large price
292: peaks and (ii) they comprise a large number of different items which
293: will play the same role as individual stocks or economic sectors in our
294: previous study.
295: \qpar
296:
297: During Word War II there was a postage stamp bubble in France which
298: was triggered by the fact that inflation was high (the consumer price
299: index increased from 130 in 1941 to 350 in 1945) while at the same time,
300: due to war restrictions there was only a small outlet for consumption.
301: Thus almost all stamps experienced a price peak (usually culminating in
302: 1944) with their real (i.e. deflated) price multiplied by a factor
303: of 3 or 4. The present investigation focuses on 19th century French stamps.
304: Table 1 shows that the correlation between peak and bottom amplitude
305: although still significant is smaller than in previous cases. This can
306: possibly be attributed to the fact that the prices given in
307: stamp catalogs are estimates made by stamp experts and traders rather than
308: real market prices. Estimates are particularly difficult to make for stamps
309: which are not traded very often; such is the case for the stamps which
310: are particularly rare and expensive. Now (see in this respect
311: [11] % Roehner and Sornette 1999
312: ) the expensive stamps are also those for which the
313: peak amplitude $ A $ is largest and any bias for large $ A $ sample
314: points will notably affect the correlation and the slope of the regression
315: line.
316: \qpar
317:
318: Our next example concerns the real estate market in Paris. Between
319: 1985 and 1995 there was a price peak which resulted in a doubling, and
320: in some areas a three fold increase, of apartment prices. First, we consider
321: the prices in 20 different districts (``arrondissements'')
322: of downtown Paris (i.e. the so-called ``Paris intra-muros''
323: area).
324: The regression leads to values
325: for the correlation and for the slope of the regression line which
326: are similar to those obtained in the case of stock markets.
327: As shown in Fig.5 the sample points in the $ (A,B) $ plane are evenly
328: distributed along the regression line and do not display any obvious
329: non-linear effect.
330: \qL
331: The second result concerns the same price peak but for apartments
332: according to their size (from one- to five-room). Not surprisingly, due
333: to the small number of sample points the error margin is fairly large;
334: note that the value $ a =0.4 $ obtained in previous cases is within the
335: error bars.
336: \qpar
337:
338: Up to that point all our results were consistent with the resilience
339: pattern, but the following case lead to an unexpected result. It
340: concerns the real estate
341: bubble which occurred in Britain in the 1980s; from
342: the area of London where it started, it spread progressively northward
343: to the rest of the country. Price results can be analyzed at the level
344: of each of the 11 regions composing Britain. The result came as a surprise;
345: in this case
346: the amplitudes are negatively correlated and the correlation
347: is fairly low. In order to see if the negative correlation is found elsewhere
348: it would be of great interest to perform a similar test for other countries;
349: unfortunately, in France reliable regional housing prices are only available
350: since 1995.
351:
352:
353: % TABLE 1
354: \begin{table}[t]
355:
356: \centerline{\bf Table 1 \ The resilience pattern: relationship between}
357: \centerline{\bf peak amplitude $ (A) $ and bottom amplitude
358: $ (B) $: $ B=aA+b $}
359:
360: \vskip 5mm
361: \hrule
362: \vskip 0.5mm
363: \hrule
364: \vskip 2mm
365:
366: $$ \matrix{
367: \tvi & \hbox{Market} & \hbox{Peak} & \hbox{Number} & a & b &
368: \hbox{Correlation} \cr
369: \tvi & \hbox{} & \hbox{} & \hbox{of items} & & & \hbox{} \cr
370: \noalign{\hrule}
371: \tvi & \hbox{Stocks} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
372: 1 & \hbox{NYSE} \hfill& 1929\hbox{ Oct} & 85 & 0.40 \pm 0.05 & -0.27 \pm 0.24 &
373: 0.87 \cr
374: 2 & \hbox{Paris}\hfill & 1929\hbox{ Feb} & 19 & 0.44 \pm 0.17 & -0.03 \pm 0.29 &
375: 0.76 \cr
376: 3a & \hbox{Tokyo}\hfill & 1989\hbox{ Oct} & 26 & 0.39 \pm 0.11 & 0.27 \pm 0.10 &
377: 0.80 \cr
378: 3b & \hbox{Tokyo}\hfill & 1989\hbox{ Oct} & 26 & 0.40 \pm 0.09 & 0.35 \pm 0.13 &
379: 0.87\cr
380: & \hbox{} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
381: & \hbox{Stamps} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
382: 4 & \hbox{France}\hfill & 1944 \hbox{} & 56 & 0.10 \pm 0.08 & 0.77 \pm 0.10 &
383: 0.33 \cr
384: & \hbox{} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
385: & \hbox{Real estate} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
386: 5 & \hbox{Paris}\hfill & 1990 \hbox{} & 20 & 0.43 \pm 0.16 & 0.73 \pm 0.04 &
387: 0.79 \cr
388: 6 & \hbox{Paris}\hfill & 1990 \hbox{} & 5 & 0.14 \pm 0.26 & 1.1 \pm 0.1 &
389: 0.52 \cr
390: 7 & \hbox{Britain}\hfill & 1989 \hbox{} &11 & -0.10 \pm 0.17 & 1.4 \pm 0.05 &
391: -0.38 \cr
392: & \hbox{} \hfill & \hbox{} & \hbox{} & & & \hbox{} \cr
393: & \hbox{Average}\hfill & \hbox{(except 7)} & & 0.32 & & 0.68 \cr
394: } $$
395:
396: \hrule
397: \vskip 0.5mm
398: \hrule
399: \vskip 5mm
400:
401: {\small Notes: All prices used in the regression are real
402: (i.e. deflated) prices.
403: Case (3a) refers to $ t_1=1985 $, while case (3b) refers to $ t_1=1980 $.
404: For some reason yet to be understood the housing bubble in Britain does
405: not follow the resilience pattern (negative correlation).
406: \qL
407: {\sl Sources: 1: [1]; 2: Annuaire Statistique de la
408: France, R\'esum\'e R\'etrospectif (1966, p.541); 3a,b: Japan
409: Statistical Yearbook (various years); 4: [5]; 5,6: Conseil par des notaires
410: (23 Dec. 1991) and Chambre des Notaires; 7: Halifax index.}}
411:
412: \end{table}
413:
414:
415:
416: \qI{Conclusion}
417:
418: About 60 percent of the 222 sample points in Table 1 concerned stock
419: markets. Thus, the evidence supporting the resilience pattern
420: for stock markets was particularly strong. Yet, it was important also
421: to show that the resilience effect is {\it not} confined to stock
422: markets for it suggests that a possible theoretical framework should
423: apply to other speculative markets as well. Let us now briefly
424: discuss the significance and possible implications of the present finding.
425: \qL
426: In the late 1990s econophysicists along with some economists ([3])
427: devoted great attention to the
428: statistical analysis of stock market indexes, the overall objective
429: being the identification of possible scaling laws. Yet, indexes do not
430: give great insight into the internal mechanisms of stock markets. Such
431: an understanding can only be gained by opening the ``black box''
432: and studying the interactions that take place between individual stocks.
433: This idea has recently gained more acceptance as shown by a number
434: of innovative papers going into that direction; e.g.
435: [3, % Lux (1996)
436: 4, % Mantegna (1999)
437: 6 % Plerou et al. (2000)
438: ].
439: \qL
440: Finally, let us briefly consider the next step, namely the construction
441: of a theoretical framework. Obviously any model is
442: (and has to be) a schematization of the real world; therefore,
443: constructing a ``realistic'' model cannot be a viable and suitable
444: objective; models need more precise ``targets'' and ``guiding lights''.
445: In a number of recent empirical studies we have tried to define
446: such targets: the sharp peak - flat trough pattern
447: [9, 10], % Roehner (2001), Roehner and Sornette (1998)
448: the price multiplier effect
449: [7, 9], % Roehner (2001), Roehner (2000)
450: the relationship between stock market crashes and increases in interest
451: rate spread
452: [8] % Roehner (2000)
453: define quantitative patterns which provide useful guiding lights
454: for the construction of a theoretical framework.
455: \qL
456: On the theoretical side some promising advances have been made recently
457: which can possible provide an adequate framework for the description of the
458: internal machinery of stock markets. For instance one would not be
459: surprised to see percolation (see in this respect
460: [2]) % Goldenberg (2000)
461: play a role in the spread of a bubble;
462: after all, a speculative outburst
463: can be seen as propagating from high-growth stocks
464: to low-growth stocks in the same way as a technical
465: innovation progressively
466: gains acceptance in an economy.
467: \qpar
468:
469: {\bf Acknowledgments}\quad I would like to express my gratitude to the
470: statistical experts of the Halifax Company (U.K.)
471: and the Chambre des Notaires
472: (Paris) for their kind assistance.
473:
474:
475:
476:
477:
478: \vfill \eject
479:
480:
481: \centerline{\bf \Large References}
482:
483: \vskip 1cm
484:
485:
486: \qparr
487: (1) COMMON STOCK price histories 1910-1986 and logarithmic supplement, 1988:
488: WIT Financial Publishers. Anchorage.
489:
490: \qparr
491: (2) GOLDENBERG (J.), LIBAI (B.), SOLOMON (S.), JAN (N.), STAUFFER (D.) 2000:
492: Marketing percolation. Physica A (to appear). Cond-mat/0005426.
493:
494: \qparr
495: (3) LUX (T.) 1996: The stable Paretian hypothesis and the frequency of
496: large returns: an examination of major German stocks.
497: Applied Financial Economics 6,463-475.
498:
499:
500: \qparr
501: (4) MANTEGNA (R.N.) 1999: Hierarchical structure in financial markets.
502: European Physical Journal B 11,1,193-197.
503:
504: \qparr
505: (5) MASSACRIER (A.) 1978: {\sl Prix des timbres-poste classiques de 1904 \`a 1975.}
506: A. Maury. Paris.
507:
508: \qparr
509: (6) PLEROU (V.), GOPIKRISHNAN (P.), AMARAL (L.A.N.), MEYER (M.), STANLEY (H.E.)
510: 1999: Scaling of the distribution of price fluctuations of individual
511: companies. Physical Review E 60,6,6519-6529.
512:
513: \qparr
514: (7) ROEHNER (B.M.) 2000: Speculative trading: the price multiplier effect.
515: The European Physical Journal B 14,395-399.
516:
517: \qparr
518: (8) ROEHNER (B.M.) 2000: Identifying the bottom line after a stock market
519: crash. International Journal of Modern Physics 11,1,91-100.
520:
521: \qparr
522: (9) ROEHNER (B.M.) 2001: {\sl Hidden collective factors in speculative trading.}
523: Springer-Verlag. Berlin (to appear).
524:
525: \qparr
526: (10) ROEHNER (B.M.), SORNETTE (D.) 1998: The sharp peak - flat trough pattern
527: and critical speculation. The European Physical Journal B 4,387-389.
528:
529: \qparr
530: (11) ROEHNER (B.M.), SORNETTE (D.) 1999: Analysis of the phenomenon of
531: speculative trading in one of its basic manifestations: postage stamp
532: bubbles. International Physical Journal B (to appear).
533:
534: \qparr
535: (12) TILLY (C.) 1993: {\sl European revolutions 1492-1992.}
536: Blackwell. Oxford.
537:
538:
539: \vfill \eject
540:
541: % Imprimante recto-verso et decoupage
542: %\def\qpage{\vfill \eject}
543: \def\qpage{\vfill \eject \ \vfill \eject}
544:
545: \def\qv{\vskip 0.1mm plus 0.05mm minus 0.05mm}
546: \def\qhu{\hskip 0.6mm}
547: \def\qhv{\hskip 3mm}
548: \def\qhw{\hskip 1.5mm}
549:
550: %\parindent=1.5cm \narrower
551:
552: \baselineskip=11pt
553:
554: \def\qleg#1#2#3{ \noindent
555: {\bf \small #1\qhw}{\small #2\qhw}{\it \small #3}\qv }
556:
557: \qleg{Fig.1\qhv Course of the price for three stocks versus Dow Jones
558: Index.}
559: {The prices are yearly highs. During the bull market 1921-1929, the
560: prices of Coca Cola and Columbia Gas (an utility company) have
561: increased faster and more than the DJ average, while the price of
562: Burlington Northern (a railroad company)
563: has increased slower and less than the DJ.
564: The circles and squares show the peaks and troughs respectively.
565: The resilience pattern reveals itself in the fact that when a peak is above
566: (or below) the DJ average the same situation prevails for the
567: trough.}
568: {Sources: Common stock price histories (log supplement); Dow Jones
569: Investor's handbook (1972).}
570:
571: \qleg{Fig.2\qhv Parallel between the crashes of 1929 and 1987.}
572: {Monthly prices. During the 8 months before and after the crash
573: the behavior of the Dow Jones Index was very much the same in
574: both cases; as one knows the subsequent evolution was very different
575: however. This suggests that crashes and long-lasting slides are
576: two distinct (and not necessarily related) phenomena.}
577: {Sources: The Dow Jones averages 1885-1970; OECD main economic
578: indicators (1969-1988).}
579:
580: \qleg{Fig.3\qhv NYSE, 1921-1932 peak: distribution of 85 sample
581: points in the $ (A,B) $ plane.}
582: {Each triangle corresponds to a stock;
583: the prices are yearly highs. Dotted line: linear regression (correlation
584: is 0.87).}
585: {Source: Common stock price histories (log supplement).}
586:
587: \qleg{Fig.4\qhv Long-term behavior of the price of Columbia Gas System.}
588: {The prices are yearly highs.
589: The price of that utility stock increased tremendously during the
590: 1921-1929 bull market; then, from 1932 to 1936 it fluctuated around
591: 200 before falling to about 30; it never really recovered in the
592: second half of the 20th century. The chart does not cover the
593: 1990s because there was a change in the name of the company; however
594: it can be noted that during the 1990-2000 bull market the Dow Jones
595: Utilility index increased 1.7 times less than the Standard and Poor's 500;
596: this would give a stock price of about 200 in year 2000, still well
597: below the 1929 high.}
598: {Source: Common stock price histories (log supplement).}
599:
600: \qleg{Fig.5\qhv Paris, 1985-1995 real estate peak: distribution of 20
601: sample points in the $ (A,B) $ plane.}
602: {The numbers correspond to the 20 arrondissements.}
603: {Source: Chambre des Notaires.}
604:
605:
606:
607: \end{document}
608: