Investors, Information Arrivals, and Market Liquidity: Empirical Evidence from Financial Markets
|Kustantaja||Tampere University of Technology|
|Tila||Julkaistu - 19 lokakuuta 2018|
|Nimi||Tampere University of Technology. Publication|
Earlier research on these topics exists, but the interrelations between these factors have not been studied in depth. The objective of this thesis is to improve our knowledge of the interrelations between investors, information arrivals, and liquidity in the context of ﬁnancial markets. By addressing several research gaps related to these themes, this thesis aims to provide new empirical evidence in order to help the scientiﬁc community develop more reliable and robust models that describe the markets; in general, a better understanding of these topics and such interrelations may help improve regulations, exchange organizations, and investment management.
This thesis consists of an introductory part and four research papers (Articles I – IV). Article I uses logistic regression to study how Nokia’s Facebooks posts and related activities are associated with investors’ decisions to buy versus sell Nokia stock. In Article II, a framework from event studies is combined with high-frequency limit order book data to examine how liquidity in stock limit order books evolves around scheduled and nonscheduled company announcements. Article III applies regression analysis to identify the factors aﬀecting the magnitude of order book liquidity shocks that company announcement releases cause in the limit order books. Finally, Article IV uses a unique data set to study the proportion of liquidity streams that a trader observes in a foreign exchange (FX) liquidity aggregator, as well as quantiﬁes a trader’s theoretical improvements in the observed spread and the cost savings when comparing the current situation with the optimal combination of streams; the optimal combinations are solved using a genetic algorithm (GA).
Earlier literature has studied how news articles aﬀect the trading of diﬀerent investors, and this thesis contributes by providing evidence that the (potentially biased) information a company releases on social media aﬀects the behaviors of diﬀerent investors in the stock market diﬀerently. While the decisions of arguably less sophisticated investors—passive households and non-proﬁt organizations—are associated with Facebook data, those of more sophisticated investors—ﬁnancial institutions—seem to be independent of Facebook data.
Moreover, company announcements are found to cause signiﬁcant changes in the stock limit order book liquidity, which is inconsistent with the ﬁnding of an earlier study using news data. In particular, scheduled announcement releases may improve liquidity to an abnormally high level, indicating that scheduled announcement releases resolve asymmetric information problems in the market, whereas the order book liquidity remains relatively low in many cases still an hour after the non-scheduled announcement releases The immediate liquidity shocks following the announcement releases are ampliﬁed by order book asymmetry prior to the announcements releases. Moreover, a fast reaction is a strong reaction (the faster the illiquidity peak is reached after the announcement release, the larger the peak usually is), and in case of non-scheduled announcement releases, recent losses amplify the liquidity shocks. The ﬁndings also indicate that liquidity measured over multiple order book price levels behaves quite diﬀerently compared to the conventional bid – ask spread calculated using data from the best order book levels, indicating that measuring liquidity just using top-of-the-book data may lead to misleading inferences.
Finally, the results show that in a liquidity aggregator, traders observe only a small proportion of liquidity streams available: on average, a trader observes 5.4 streams out of the total 165 streams provided by 42 liquidity providers (the maximum is 23 and the minimum is 1). However, traders observe relatively tight spreads already with four or ﬁve streams, and traders with more streams observe only marginal improvements in spread, if any. In theory, most traders could cut their observed spread by more than a half and save up to $0.18 basis points per e1 traded with the optimal combination of liquidity streams; in practice, however, traders may not be able to exploit the improvements because they are not free to choose just any streams in the aggregator, and if they would change the streams they observe, the liquidity providers would likely change their quoting behavior. Nevertheless, the novel empirical results can be used to assess the eﬃciency of the aggregator as a trading technology and the liquidity provision in the FX market, in general.