Abstract
Using data on over 18 million trades from the New Zealand Stock Exchange (NZX), this paper examines how market segmentation affects overall market quality in a market that until recently had no restrictions on trading outside the central limit order book (LOB). We find that upstairs trading results in lower transaction costs, larger trade size and lower volatility. A newly implemented minimum size requirement for trades in the upstairs market has the desired outcome and further lowers transaction costs and volatility due to reduced market segmentation. The results suggest that a functional upstairs market can have a positive impact on market quality if well-designed.
Acknowledgements
We thank participants at the NZX for helpful comments during a seminar presentation in January 2021 and for feedback received at the 25th New Zealand Finance Colloquium (NZFC) 2021. Research support (access to supercomputer) from the UC Research Compute Cluster (RCC) is also gratefully acknowledged.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 For example, the New York Stock Exchange (NYSE), London Stock Exchange (LSE), Toronto Stock Exchange (TSE), NASDAQ Nordic (OMX).
2 Trades with price improvement are exempted. See NZX Participant Rules 10th December 2020 section 13.2.1.
3 Based on a review of historical data, the NZX estimated that around 5–7% of upstairs trades would be affected by the new rules (see NZX Participant Rule Consultation released in April 2018).
4 The NZX50 represents about 90% of the total market capitalisation.
5 To match the sample period after the policy change, Panel A is based on data over 18 months before the minimum trade size for upstairs trades came into effect (April 2017 to September 2018).
6 Smaller trades in the LOB market are also driven by an increase in retail participation at the NZX over time.
7 We thank an anonymous reviewer for suggesting to examine simultaneous effects between the variables in a multivariate analysis. Estimating independent OLS regressions as in prior literature (e.g. Rose, Citation2014) provides similar results.
8 Coefficients on the lagged variables are not reported for brevity, but the variables are described by several of their own and lags of the other variables.
9 We report results based on a sub-sample of about one and a half years on either side of the policy change. However, we find virtually identical results using data over the entire sample period starting in September 2010.
10 The negative coefficients on the policy dummy and interaction terms in case of the aggregate market are mainly driven by downstairs trades. Using only upstairs trades, the coefficient on the policy dummy is positive and statistically significant at the 1% level, as one would expect from implementing a minimum trade size.
11 We would like to thank the NZX for suggesting this additional explanation.
12 We report results using realised volatility estimated over 60-minute intervals, but find similar results using realised volatility estimated over 15, 30 and 90-minutes.