George Sofianos

An empirical analysis of NYSE specialist trading

Abstract This paper examines empirically the magnitude and determinants of dealer trading by NYSE market makers (specialists) across stocks and over time. Across stocks, specialist dealer trading varies widely and is inversely related to trading volume and proxies for off-exchange competition. Over time in an individual stock, specialists participate more actively as sellers (buyers) when holding long (short) inventory positions. This results suggest that dealers control their inventory positions by selectively timing the size and direction of their trades rather than by adjusting their quotes. Further, specialists participate more in smaller trades and when the bid–ask spread is wide.

Liquidity provision and specialist trading in NYSE-listed non-U.S. stocks

Abstract We examine how the intrinsic differences between U.S. and non-U.S. stocks affect market participants and the market quality of non-U.S. stocks relative to U.S. stocks. Using proprietary data on NYSE specialist trading, we find that, all else equal, specialist closing inventory positions for non-U.S. stocks are closer to zero than U.S. stocks. The evidence on specialist participation and stabilization rates is mixed. Non-U.S. stocks from developed markets have higher specialist participation and stabilization rates than U.S. stocks, while emerging market stocks have lower participation and stabilization rates than U.S. stocks. With respect to market quality, we find that, all else equal, non-U.S. stocks have wider spreads, less depth, and greater transitory volatility than U.S. stocks. We investigate the reasons behind the difference in liquidity and find that the larger non-U.S. spreads are primarily due to higher information asymmetry and increased adverse selection risk. We conclude that liquidity providers demand greater compensation for trading non-U.S. stocks, but this additional compensation is necessary to offset the higher adverse selection risk.

The trades of NYSE floor brokers

Abstract This paper studies the contribution of NYSE floor brokers to the Exchanges agency auction market. Floor brokers represent 44%, specialists 11%, and system orders 45% of the value of all executed orders in our sample. We analyze how the cross-sectional distribution of floor broker trading depends on liquidity, block volume, on- and off-exchange competition, volatility, and order flow internalization. Floor brokers participate in large trades, primarily in liquid stocks, and they trade more when volatility is high. They provide two-sided liquidity to the market and often provide liquidity that would otherwise have been supplied by NYSE specialists.

Quantifying market order execution quality at the New York stock exchange

Abstract This paper provides a detailed description of how to quantify best execution on market orders. We address two important issues: how to measure best execution on market orders and the need for cross-market standardization to facilitate comparisons. We propose six measures for quantifying best execution on eligible system market orders and present NYSE estimates for August 1999. These measures include percent executions inside the quote, percent executions outside the quote, percent discount relative to the quoted price, percent executions receiving depth improvement, percent single price executions, and turnaround time. We also show that best-execution statistics are sensitive to the calculation methodology. For valid cross-market comparisons, therefore, we must standardize methodologies and control for other factors that may affect best execution performance.

Quantifying the SIGMA X Crossing Benefit: The Sequel

FALL 2008 In this report we investigate the benefits of crossing orders in dark pools. In the United States, the Goldman Sachs SIGMA smart router crosses 31% of market and marketable limit orders within SIGMA X, the Goldman Sachs pool of non-displayed liquidity. We focus on marketable orders that clients directly route through SIGMA and quantify the reduction in execution shortfall resulting from SIGMA X crossing. We isolate the SIGMA X effect by using multivariate regressions to control for all the other factors that may inf luence execution shortfall. Exhibit 1 summarizes our f indings. For order sizes greater than quoted depth, SIGMA X crossing reduces execution shortfall and the SIGMA X crossing benefit increases as order size relative to quoted depth increases. For example:

When the Going Gets Tough, the Algos Get Going

This article examines how the market turmoil of September to November 2008 affected trading costs in general performance, and algo performance in particular. It also examines how traders responded to the new challenges created by the market turmoil. How does one execute trades in the new environment? Trade more high-touch or low-touch? Trade more or less aggressively? The analysis compares the market turmoil period to a normal “benchmark” period, May 2008. It finds that in the market turmoil period: 1) quoted spreads on the S&P 500 stocks increased 50% and quoted depth decreased 50%, 2) algo usage remained steady at 21% of executed value but algo clients shifted to more aggressive executions, 3) the liquidity impact of the average 20,000-share order in the sample more than doubled to 25 bps (10 bps in May), and 4) alpha-to-close was higher, so despite the higher liquidity impact the average algo client captured some alpha-to-close.

The Impact of an Increase in Volatility on Trading Costs

The current low volatility environment keeps trading costs low. But what if volatility increases? In this paper, we discuss how an increase in volatility will affect execution strategies and estimate its impact on portfolio trading costs. We derive our estimates by simulating the Goldman Sachs PortX algorithm. PortX derives the optimum execution strategy by balancing the cost-risk trade-off and therefore allows us to examine how an increase in volatility changes the optimum strategy. Our simulations suggest that doubling volatility will increase the expected trading cost on large cap portfolios by 20% to 25% (depending on trading aggressiveness) and by almost 50% on small cap portfolios. In our simulations, doubling volatility increases trading costs both directly (for a fixed execution horizon) and indirectly: the resulting increase in execution risk reduces the optimum execution horizon further increasing trading costs.

Accessing Displayed and Non-Displayed Liquidity

U.S. equity markets are going through a period of dramatic change that is reshaping the way traders access and consolidate liquidity. In this article, we survey the state of liquidity in the U.S. equity markets, carefully distinguishing between displayed and non-displayed liquidity. Over the past decade, automation in the market for NASDAQ stocks made displayed liquidity easy to access electronically, consolidate and commoditize. Increased automation on the NYSE and the regional exchanges is about to do the same for displayed liquidity in NYSE stocks. Most liquidity, however, is non-displayed and efficient execution of large order requires accessing all types of liquidity. Electronically accessing and consolidating non-displayed liquidity is the next frontier, with several initiatives attempting to tackle the challenge.

Choosing Benchmarks vs. Choosing Strategies: Part 2 - Execution Strategies: VWAP or Shortfall

In this article, we use execution data from the Goldman Sachs algorithmic trading desk to compare the performance of VWAP and shortfall algorithms. Our analysis shows that the two algorithms work as intended: shortfall performs better than VWAP when short-term (ST) alpha is high, helping traders to reduce alpha loss. The shortfall algorithm also delivers lower execution risk. We also find, however, that the average trader does not optimally allocate orders between the VWAP and shortfall algorithms. Our findings suggest that providers should better educate users on the relative merits of VWAP and shortfall algorithms and better quantify the trade-offs. If users cannot easily differentiate between high and low ST-alpha orders, a practical alternative is to choose the one algorithm that best fits the overall characteristics of their order flow.

Execution Quality at the New, Fast NYSE

Between October 06 and January 07, the NYSE was dramatically transformed into a fast execution venue with sub-second execution times (Exhibit 1). In this report, we examine the impact of this transformation on the cost of trading at the NYSE. We find that the NYSE achieved the sharp drop in execution times without major changes in trading costs, at least for executions of up to 10,000 shares.