Student thesis: Phd


The main intellectual contribution of this thesis consists of three self-contained essays. These essays are preceded by an introduction and literature review (chapters 1 and 2), and their results summarised in a concluding chapter 6. In the core essays, which constitute chapters 3-5 inclusive of the thesis, I test the empirical implications of two market microstructure invariance (MMI) principles first proposed by Kyle and Obizhaeva (2016b). The market setting I choose analyses trades in FTSE 100 index constituent stocks for the period between January 2007 and December 2009, a period which incorporates both the 2008-09 financial crisis and the introduction of alternative trading platforms for FTSE 100 stocks The chapter 3 examines the MMI of bets, as applied to trades, in FTSE 100 index constituent stocks. To link bets and trades the thesis formulates an extended version of ITI model by Andersen et al. (2018) motivated by the MMI model (Kyle and Obizhaeva, 2016b). The model accounts for the level of trade intermediation and order shredding. I empirically test the model’s trading activity prediction on trade data using panel estimation methodologies. I find that for highly capitalized stocks, trade counts yield the predicted 2/3 proportionality relationship to trading activity. Further investigations, using alternative notions of trading activity proposed by Clark (1973) and Ané and Geman (2000) reveal the predicted proportionality only for large tradesize stocks. The chapter 4 develops the analysis of the earlier chapter by investigating the market microstructure invariance proposition in FTSE 100 constituent stocks at the level of individual stocks, using four different notions of trading activity. I find that the notion of trading activity proposed by Clark (1973) reveals the predicted 2/3 proportionality number for the majority of stocks. This result is consistent even when the first and last minutes of active trading in the market are excluded. Invariance models yield a 1/2 proportionality between the log values of trade counts and trading activity, whereas the intraday trading patterns in the specific market, the magnitude of trade size and its correlation with the volatility partly explain this value. Based on this, I show that analysis on a year by year, pre-crisis and in-crisis sample do not suggest a unified order flow composition across stocks. The chapter 5 focuses on the MMI of transaction costs. I empirically test the respective predictions using three common proxies for transaction costs, namely quoted, effective, realized spreads on FTSE 100 stock trading data. As predicted by market microstructure invariance, I find that a -1/3 proportionality is present in average daily patterns in our sample for all three proxies of transaction costs, with larger trades having a negative impact on this proportionality when the underlying variables are estimated as intraday averages. My results suggest that market fragmentation does not impact the estimated invariance coefficients, though trading activity and volume traded on alternative platforms are negatively correlated with the percentage transaction costs on LSE per unit of volatility. Finally, the invariance prediction holds for a consolidated market, but the lower reduction in the realised spreads may suggest a greater impact of large trades in the alternative platforms or the fact that only few market makers benefit from an increase in trading activity.
Date of Award1 Aug 2019
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorMichael Bowe (Supervisor), Stuart Hyde (Supervisor) & Sarah Zhang (Supervisor)


  • Bid-ask spread
  • FTSE 100 equities
  • Transaction costs
  • Market impact
  • Market Fragmentation
  • Order Size
  • Microstructure
  • Intraday Trading Invariance
  • Market Microstructure Invariance
  • Liquidity

Cite this