We study what properties a liquidity surface (LS) must possess when the supply and demand of a given market are in equilibrium. To formalize this concept, we first analyze foreign exchange markets, where symmetry between the buy and sell sides just amounts to the invariance of market impact with respect to a change of base currency. We speak in this case of supply–demand symmetry, and call the corresponding LS symmetrical. We then extend the idea to general securities, applying similar change of numeraire arguments. We give necessary and sufficient conditions for characterizing symmetrical LS's and we classify all possible solutions to these conditions, providing some stylized examples. Symmetrical LS's are interpreted as equilibria around which liquidity imbalances fluctuate. We show that market impact functions which are even with respect to order size – usually adopted by market impact models in the literature to represent bid–ask symmetry – are not symmetrical LS's and are always associated to an excess of supply; they are approximations of symmetrical LS's only in the limit of highly liquid regimes (no bid–ask spread, small orders, small market impact).
| Citation: |
Figure 2. Graphical interpretation of supply–demand symmetry. A regular LS is symmetrical if and only if the plot of $ y = L(x)/m $ is symmetrical with respect to $ y = -x $, concave, increasing and passing through the origin. The orange line is such an example. These curves are confined to stay in the white area of the plane. Two extreme cases exist: the green line, which represents a perfectly liquid market, where quotes at all sizes coincide with fair price, ($ m(s) = m $, $ \forall s\in\mathbb{R} $); and the red line, which is the most illiquid possible market, where there aren't any quotes either on the bid or on the ask side ($ m(s) = +\infty $ if $ s<0 $, $ m(s) = 0 $ if $ s>0 $)
Figure 6. In the top plot the MSDC and SDC of the example in Figure 5. In the bottom plot, the corresponding marginal and average impacts. In this example, quotes decay to zero on the bid wing and diverge to infinity at some finite ask size
Figure 8. In the top plot the MSDC and SDC of the example in Figure 7. In the bottom plot, the corresponding marginal and average impacts
Figure 9. An asymptotically linear symmetrical Liquidation Operator $ L(s) $ with zero bid-offer spread. The corresponding MSDC is asymptotically flat as illustrated in Figure 12
Figure 10. In the top plot the MSDC and SDC of the example in Figure 9. In the bottom plot, the corresponding marginal and average impacts. MSDC and SDC are asymptotically flat in this example. These plots, at small size scales, are close to even functions in the sense of equation (1)
Figure 11. An asymptotically linear symmetrical Liquidation Operator $ L(s) $ with finite bid-offer spread. The corresponding MSDC is asymptotically flat as illustrated in Figure 12
Figure 12. In the top plot the MSDC and SDC of the example in Figure 11. In the bottom plot, the corresponding marginal and average impacts
Figure 14. In the top plot the MSDC and SDC of the example in Figure 13. Both the MSDC and the SDC are linear on the ask side. The bid side is clearly not linear
Figure 16. In the top plot the MSDC and SDC of the example in Figure 5. Both the MSDC and the SDC are linear on the ask side. The bid side is clearly not linear
Figure 17. A symmetrical Liquidation Operator $ L(s) $ in the case where there are four finite bid quotes and four symmetrical finite ask quotes. The corresponding MSDC is piecewise constant, as illustrated in Figure 18
Figure 18. In the top plot the MSDC and SDC of the example in Figure 17. In the bottom plot, the corresponding marginal and average impacts. The visualization of the symmetry relation between corresponding bid and ask quotes is all but intuitive. Symmetrically placed bid and ask quotes have size and price in correspondence with each others as described by Proposition 3.6
Figure 20. In the top plot the MSDC and SDC of example in Figure 19. In the bottom plot, the corresponding marginal and average impacts. This example, as compared with the one in Figure 18, is made of a strip of finer quotes and gives rise to a more regular pattern, with a behavior at small size scales which is closer to an even function
Figure 21. In the top plot a symmetrical LS obtained from a power–law (square–root) MSDC on the bid side and no bid–offer spread. The marginal impact is shown in the bottom plot to display that for small degree of illiquidity and/or small size, impact can be approximated by an even function, as proved in Corollary 3.9
Figure 22. A similar example to the one in Figure 21 but with a finite bid–offer spread. The bottom plot shows impact as a difference from mid–price. It can be noticed that in this case, impact is not an even function even at small size scales, as described in Corollary 3.9
| [1] |
C. Acerbi and Zs. Szekeres, Introduction to LiquidityMetrics, MSCI Technical Document, 2013.
|
| [2] |
A. Alfonsi, A. Fruth and A. Schied, Optimal execution strategies in limit order books with general shape functions, Quant. Finance, 10 (2010), 143-157.
|
| [3] |
R. Almgren and N. Chriss, Optimal execution of portfolio transactions, The Journal of Risk, 3 (2001), 5-39.
doi: 10.21314/JOR.2001.041.
|
| [4] |
R. Almgren, C. Thum, E. Hauptmann and H. Li, Equity market impact, Risk Magazine, 2005.
|
| [5] |
BARRA, Market Impact Model Handbook, 1997.
|
| [6] |
J.-P. Bouchaud, J. D. Farmer and F. Lillo, How markets slowly digest changes in supply and demand, working paper, available at "ssrn.com", 2008.
doi: 10.2139/ssrn.1266681.
|
| [7] |
J.-P. Bouchaud, Y. Gefen, M. Potters and M. Wyart, Fluctuations and response in financial markets: The subtle nature of 'random' price changes, Quantitative Finance, 2004.
doi: 10.2139/ssrn.507322.
|
| [8] |
J. Gatheral, No-dynamic-arbitrage and market impact, Quantitative Finance, 10 (2010), 749-759.
doi: 10.1080/14697680903373692.
|
| [9] |
E. Goursat, A course in mathematical analysis, Vol. I, Ginn and Company, (1904), 404-405. https://archive.org/details/dli.ernet.523782/page/404/mode/2up.
|
| [10] |
G. Huberman and W. Stanzl, Price manipulation and quasi-arbitrage, Econometrica, 72 (2004), 1247-1275.
doi: 10.1111/j.1468-0262.2004.00531.x.
|
| [11] |
J. L. Lagrange, Nouvelle méthode pour résoudre les équations littérales par le moyen des séries, Mémoires de l'Académie Royale des Sciences et Belles–Lettres de Berlin, 24 (1770), 251-326.
|
| [12] |
A. Obizhaeva and J. Wang, Optimal trading strategy and supply-demand dynamics, Discussion Paper MIT Sloan School of Management, 2005.
|
The market impact of an hypothetical currency pair, which is even when expressed in one currency (blue line) is not even when expressed in the opposite currency (red line)
Graphical interpretation of supply–demand symmetry. A regular LS is symmetrical if and only if the plot of
Graphical interpretation of the conjugation relationship
Illustration of Proposition 3.16.
A symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 5. In the bottom plot, the corresponding marginal and average impacts. In this example, quotes decay to zero on the bid wing and diverge to infinity at some finite ask size
A symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 7. In the bottom plot, the corresponding marginal and average impacts
An asymptotically linear symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 9. In the bottom plot, the corresponding marginal and average impacts. MSDC and SDC are asymptotically flat in this example. These plots, at small size scales, are close to even functions in the sense of equation (1)
An asymptotically linear symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 11. In the bottom plot, the corresponding marginal and average impacts
A symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 13. Both the MSDC and the SDC are linear on the ask side. The bid side is clearly not linear
A symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 5. Both the MSDC and the SDC are linear on the ask side. The bid side is clearly not linear
A symmetrical Liquidation Operator
In the top plot the MSDC and SDC of the example in Figure 17. In the bottom plot, the corresponding marginal and average impacts. The visualization of the symmetry relation between corresponding bid and ask quotes is all but intuitive. Symmetrically placed bid and ask quotes have size and price in correspondence with each others as described by Proposition 3.6
Another symmetrical Liquidation Operator
In the top plot the MSDC and SDC of example in Figure 19. In the bottom plot, the corresponding marginal and average impacts. This example, as compared with the one in Figure 18, is made of a strip of finer quotes and gives rise to a more regular pattern, with a behavior at small size scales which is closer to an even function
In the top plot a symmetrical LS obtained from a power–law (square–root) MSDC on the bid side and no bid–offer spread. The marginal impact is shown in the bottom plot to display that for small degree of illiquidity and/or small size, impact can be approximated by an even function, as proved in Corollary 3.9
A similar example to the one in Figure 21 but with a finite bid–offer spread. The bottom plot shows impact as a difference from mid–price. It can be noticed that in this case, impact is not an even function even at small size scales, as described in Corollary 3.9