Multi-Factor Stock Selection for Hong Kong Stock Connect: Building a Practical Quant Model

Introduction: Why Hong Kong Stock Connect Needs a Different Investment Approach

In recent years, the Stock Connect mechanism has made it much easier for mainland Chinese investors to participate in the Hong Kong stock market. However, the Hong Kong market is quite different from the A-share market in terms of trading rules, liquidity, and investor structure. If you simply copy the stock selection methods used in A-shares, they often don't work well in Hong Kong. So, we need a quant model that fits the Hong Kong market. This article will explain in a simple way how to build a practical multi-factor stock selection model for Stock Connect stocks. We'll talk about the "Hong Kong multi-factor model" and also share some insights on "profitability in Hong Kong stocks" and "stable Hong Kong stock strategies."

Step 1: Remove Unreliable Stocks First

Why the Sample Pool Matters

In the Hong Kong market, there are many low-priced, thin-traded stocks known as "penny stocks." These stocks are easy to manipulate and their fundamental data is often unreliable. So, the first step is to filter out these bad apples.

How to Create a Clean Sample Pool

You can use a few simple conditions to filter stocks:

• Price filter: Exclude stocks with an average price below HKD 1 over the past 60 trading days.
• Liquidity filter: Exclude stocks with an average daily turnover of less than HKD 10 million over the past 20 days. Otherwise, trading costs could be high and execution difficult.
• Listing time filter: Exclude stocks listed for less than one year to avoid abnormal volatility from new listings.
• Suspension filter: Exclude stocks that are suspended for a long time or frequently suspended.

After this filtering, you will typically have around 300 to 500 mid- and large-cap quality stocks. With this clean sample pool, your subsequent analysis becomes more reliable.

Step 2: How to Build the Multi-Factor Model

What Is a Multi-Factor Model?

Simply put, a multi-factor model uses multiple indicators to score each stock, then you buy the highest-scoring ones. There are four common factor categories: Value, Quality, Momentum, and Dividend. Each category contains specific metrics.

Value Factor: Find Cheap Stocks

The value factor looks for stocks that are undervalued. Common indicators include:

• P/E ratio (PE): Lower is usually cheaper.
• P/B ratio (PB): Also, lower is usually cheaper.
• P/CF ratio (PCF): Based on operating cash flow, this is harder to manipulate.

In Hong Kong, you can also add dividend yield as an extra value dimension, since many Hong Kong companies pay decent dividends.

Quality Factor: Find Companies with Stable Earnings

The quality factor focuses on profitability and financial health. For example:

• ROE (Return on Equity): Use a three-year average to smooth out lumps.
• Gross margin stability: Companies with less volatile margins are preferred.
• Debt-to-asset ratio: Companies with lower leverage get higher scores.
• Cash earnings ratio: Operating cash flow divided by net profit. This helps avoid "paper profits."

Momentum Factor: Find Stocks in an Uptrend

The Hong Kong market shows a relatively strong trend effect. The momentum factor looks at past performance. A typical approach is to use the 12-month return, but exclude the most recent month to avoid short-term reversals. This captures the upward trend.

Dividend Factor: Earn from Dividends

High-dividend strategies have been effective in Hong Kong for a long time. Three key measures are:

• Dividend yield: The trailing 12-month yield.
• Payout ratio: Ideally between 30% and 70%. Too high or too low may be penalized.
• Dividend growth: Companies that have paid rising dividends for three consecutive years get a bonus.

Factor Synthesis: Combine the Scores

After standardizing each factor, you need to combine them into one total score. Common methods are:

• Equal-weight synthesis: Simply average the scores of the four factor categories. This is simple and robust.
• IC-weighted synthesis: Dynamically assign weights based on each factor's information coefficient (IC) over the past 12 months. Higher IC means higher weight.

It's suggested to start with equal-weight synthesis as a baseline, then test if IC-weighting improves performance. The IC-weighted approach requires a rolling window, preferably at least 36 months, to avoid excessive weight fluctuation.

Step 3: Backtesting and Rebalancing Frequency

Backtesting Is Key to Validation

Backtesting simulates your investment process using historical data to see if the strategy makes money. Let's set the parameters:

• Select the top 30 stocks with the highest composite scores at each rebalance.
• Fees: 0.1% per side, slippage at 0.2% of traded value.
• Benchmark: Hang Seng Index or Hang Seng Stock Connect Index.
• Rebalance frequencies: monthly, bi-weekly, and quarterly.

Monthly Rebalancing: The Balanced Choice

Monthly rebalancing is the first choice for many quant models. Advantages include:

• It captures factor signal changes in time.
• Turnover is moderate, so trading costs are manageable.
• It is more stable than bi-weekly rebalancing because it filters out some noise.

Bi-Weekly Rebalancing: Faster but More Expensive

Bi-weekly rebalancing reacts faster, which can be good for short-term momentum strategies. But the drawback is significantly higher trading costs. In Hong Kong, stamp duty, transaction fees, and slippage add up and can eat into profits. Data shows that bi-weekly rebalancing can drive annual turnover above 800%, with trading costs dragging net returns by 2-3%.

Quarterly Rebalancing: Cost-Efficient but Slow

Quarterly rebalancing has the lowest trading costs, but it can miss key turning points in factor performance. If factors decay quickly, quarterly rebalancing will lag behind and the strategy could suffer.

Conclusion: Monthly Wins

After accounting for reasonable trading costs, monthly rebalancing usually offers the best risk-adjusted returns (e.g., Calmar ratio). Therefore, for the Hong Kong multi-factor model, the baseline rebalancing frequency is monthly.

Backtesting Tips

Make sure the backtest period covers at least one full bull-bear cycle, for example from 2015 to 2024. Also, do out-of-sample tests—use the first 70% of data for training and the last 30% for validation—to avoid look-ahead bias. Finally, test different slippage levels (0.1%, 0.2%, 0.3%) to see how sensitive the strategy is to cost assumptions.

Step 4: Be Aware of Risks and Avoid Pitfalls

Slippage Risk

Hong Kong stocks can have very different liquidity levels. Even though we filter out low-liquidity stocks, mid- and small-cap stocks may still experience huge slippage in extreme market conditions. How to handle it?

• Set participation limits: each order should be no more than 5% of the stock's average daily volume over the past 20 days.
• Build positions in batches on rebalancing days to avoid market impact.
• Run stress tests: assume slippage is five times larger and see if the strategy remains profitable.

Overfitting Risk

Multi-factor models are prone to overfitting—finding parameter combinations that look perfect in historical data but fail in the future. Red flags include:

• Strategy performance swings wildly when factor weights or thresholds change slightly.
• The strategy performs exceptionally well only during a specific period (like a bear market) but is mediocre otherwise.
• Adding a factor dramatically boosts backtest returns, but out-of-sample performance drops sharply.

How to control overfitting?

• Keep factors simple: retain 2-3 lowly correlated sub-factors per category.
• Apply regularization (like L1 or L2 penalties) to penalty terms when optimizing factor weights.
• Use multiple cross-validation sets instead of relying on one time period.

Also, before going live, run the strategy on a simulated account for at least three months to verify its real-world feasibility.

Conclusion: A Stable Hong Kong Stock Strategy Requires Discipline

Stock Connect has opened up a broad market for mainland investors, but building an effective multi-factor model for Hong Kong stocks is not easy. In this article, we have walked through the entire practical process: sample pool construction, factor synthesis, backtesting, rebalancing, and risk control. Key takeaways are:

• Remove penny stocks and illiquid names to keep trading costs low.
• Use value, quality, momentum, and dividend factors together, with equal-weight synthesis as a robust baseline.
• Monthly rebalancing strikes the best balance between cost and timeliness.
• Backtests must include slippage and participation limits, and watch out for overfitting.

Quantitative models are attractive because they bring discipline and repeatability. But markets are always changing. We suggest that investors update factor data quarterly, re-evaluate factor effectiveness, and stay sensitive to abnormal fluctuations. Only then can a Stock Connect quant model continue to generate excess returns through both bull and bear markets. Remember, investing carries risks, and quant models are simply tools. Your own careful decision-making is what ultimately matters. We hope this article helps you invest in Hong Kong stocks with more confidence.