To begin the analysis, we present the average market shares and market quality measurements during the pre-and post-period of the SI tick size regime for our treated and control group of stocks. Then, to evaluate our hypotheses and examine if there are any significant effects in market share changes and market quality impacts, we perform a difference-in-difference regression analysis.
4.4.1. Sample period restrictions
To examine the impacts of the SI tick size regime we divide our sample period into an initial pre-and post-period between April 6, 2020, to September 25, 2020. The period covers approximately three months before ESMA enforced the tick size regime for SIs on June 26, 2020, and three months after. As Figure 3 illustrates, the pre-period begins on April 6 and ends
on June 18 (the exchange was closed June 19, 2020, due to the Swedish midsummer holiday) about a week before the event date on June 26, 2020, and covers 50 trading days. The post-period ranges between June 29 and September 25, which covers 65 trading days. Since the market share data from Fidessa Fragulator is aggregated on a weekly basis, weeks overlapping from Match and October are omitted. Furthermore, as the implementation of the tick size regime took place the last Friday in June, we omit days just before the event date, on June 26, 2020, since those days are not distinguishable within the weekly data from Fidessa.
Figure 3. Sample period
The three-month period before the event is chosen from April to control for market volatility during the spring related the coronavirus outbreak and uncertainties of the impacts on European financial markets. According to Gresse (2017), the best methodology for a pre-and post-comparison is to use consecutive months before and after the event date. Hence, we choose the three following months after June 26, 2020. However, our post-period consists of months that have historically lower trading volume due to summer vacations (Hong and Yu, 2009).
Therefore, we utilise 15 more trading days in the post-period to even out the distribution of trading volume.
While the solid braces in Figure 3 represent our two main sample periods we also utilise an alternative post-period marked by the dashed curly braces as a robustness check. The alternative post-period, which ranges from August 17 to September 11, aims to control for the lower trading volume observed during July and August in our setting. A limitation in our study is that we cannot control for effect outside of the general summer months, when the trading volume might be considered more representative in comparison to our pre-period, as the data from Nasdaq HFT is only available up to September 16. 2020.
4.4.2. Difference-in-difference regression
Our pre- and post-comparison includes the six key variables described in section 4.2 as outcome variables of interest to test if there are any significant effects from the SI tick size regime on market composition and quality. A difference-in-difference (DiD) approach allows us to control for factors such as additional regulations or seasonal events since our sample consists of stocks divided into a treated and control group. To the best of our knowledge, the SI tick size regime was a single event on June 26, which better isolates if an enforced tick size for SIs had any effects on market composition and quality at Nasdaq Stockholm. In its essence, DiD is a robust but simple and effective econometric specification to use in isolating an effect when dealing with before and after events (Bertrand et al., 2004; Foley et al., 2020).
The idea behind the DiD approach is to set up a specification using two periods (before and after the SI tick size regime) with two groups (one treated group affected by the regime and one control group that is not affected). The method is only valid under one important assumption that changes in the outcome variable over the time trend would have been the same for each group in the absence of the policy implementation, i.e., the SI tick size regime change in our case (Bertrand et al., 2004). However, since it is difficult to separate stocks that are not affected by the SI tick size regime, we approximate the difference between the groups based on SIs’
trading behaviour. Separating stocks based on their tick size is a common praxis to use (e.g., Foley et al., 2020; O’Hara et al., 2018). The market quality estimates for the treated and the control group, before and after the SI tick size regime, is compared using the DiD-regression model:
.#$ = 01)+2-+&#(3"-$+ 5#$6 + 7# + 8$+ 9#$, (6) where the outcome variable of interest .#$, for stock i in time period t, is specified as our key variables of market share and quality defined in section 4.2. 0 captures the SI tick size regime's effect. 5#$ is a set of explanatory variables that we control for similar to previous studies (e.g., Aramian and Nordén, 2021b; Foley and Putniņš, 2016; O’Hara et al., 2018; Johann et al., 2019):
volatility (logarithm of the quotient of the highest and lowest price), the logarithm of market capitalisation and the logarithm of total amount traded in the stock measured in SEK. 7# is an individual fixed effect that captures the potential heterogeneity that influences .#$ cross-sectionally but not over time, e.g., the stock's industry. 8$ is a time fixed effect that captures heterogeneity that influences .#$ over time but not cross-sectionally. Including both individual
and time-fixed effects is a common practice in the literature (e.g., Foley et al., 2020; Kwan et al., 2015). 1)+2-+&# takes a value of 1 if the stock belongs to the treatment group and 0 otherwise. (3"-$ is our time-specific binary variable that takes a value of 1 for the post-SI tick size regime period, equal or greater than June 26, and the value 0 for the pre-SI tick size regime period.