1. Discuss Numeric's investment strategy.

What is the rationale for such a strategy for Numeric's

investors?

Numeric uses quantitative models to pick stocks based on two key ideas:
 Momentum: Stocks that have positive earnings surprises and analyst upgrades tend to keep
rising.
 Value: Stocks priced below their "fair value" (based on fundamentals like earnings/book value)
tend to revert upward.
They run both long-only and long/short strategies, combining momentum and value models into a
composite score for each stock. It helps investors earn alpha (outperformance) by acting on repeatable
patterns in investor behavior—like analyst herding or delayed price reactions. Long/short strategies,
especially equitized ones, offer higher alpha (from both longs and shorts) while still giving market
exposure.

2. What is the difference between long-only and long-short strategies? Do they involve the same level of
risk?

Feature Long-Only Long-Short

Market
Always exposed to market (beta) Can be market-neutral or equitized
exposure

Strategy Buys expected winners Buys winners, shorts losers

Flexibility Benchmark-aware (sector-neutral) No benchmark constraints

Risk Simpler, but misses shorting opps Shorting adds leverage, squeeze risk

Long-short strategies aim to be market-neutral by balancing long and short positions. This allows
returns to come solely from stock selection instead of market movements. However, they come with
added risks like shorting risk, including short squeezes, and exposure to leverage, which can increase
both gains and losses. On the other hand, long-only strategies are simpler, easier to manage, and focus
on a benchmark. However, they cannot benefit from overvalued stocks. Numeric’s long-short
strategies have market-neutral versions that isolate pure alpha, along with equitized versions that add
market exposure through futures. While these strategies are more complicated and carry extra risks,
they also have the potential for higher alpha by taking advantage of both positive and negative stock
signals.

3. What is the difference between momentum and value strategies? How do they both work?

Factor Momentum Value

Idea Winners keep winning Prices revert to fair value

Inputs Earnings surprises, analyst Earnings, book value, sector,

upgrades etc.

Timeframe Short- to medium-term Medium- to long-term

Correlatio Often negatively correlated Often negatively correlated

 n

Momentum and value are two different quantitative strategies that are the basis of Numeric’s stock
selection process. Momentum strategies at Numeric rely on the idea that markets do not react enough
to new information, especially earnings surprises and changes in analyst estimates. Their momentum
model has two parts: Earnings Surprise, which captures immediate price changes in response to
earnings beats or misses, and Estrend, which tracks trends in analyst revisions. These signals are most
effective in the short to medium term, particularly right after news events.

Value strategies, in contrast, focus on the belief that prices will eventually return to their "fair value."
Numeric estimates this value using a cross-sectional regression model. This model accounts for factors
like earnings, book value, growth expectations, sector classification, and analyst coverage to determine
a stock’s statistically implied fair price. Stocks trading below that implied price are seen as undervalued
and are potential buys.

The two strategies often target opposite ends of the stock market: momentum favors recent winners,
while value targets underperformers with solid fundamentals. This approach commonly results in
negative correlation—when momentum struggles, like during market reversals, value tends to perform
well, and the opposite is true as well. Numeric takes advantage of this by using a skill-adjusted
composite model that assesses how effective each factor is for a specific stock or sector at that
moment. This flexible combination helps stabilize performance and lessen dependence on the success
of any single factor.

4. Analyze the transactions of April 29, 1997. How and why are these representatives of Numeric’s
general investment strategy?

On April 29, 1997, Numeric carried out several trades that highlighted its model-driven, data-heavy
investment strategy, especially during earnings season. The firm responded to real-time earnings
announcements by recalculating momentum and value scores and making quick trading decisions
based on analyst comments and news updates.

For example, PepsiCo announced earnings of 27 cents per share, exceeding the consensus expectation
of 24 cents. Numeric held a short position due to a negative momentum score. This surprise led to an
immediate recalculation of scores; momentum changed from -0.49 to +1.35, and the value score also
turned slightly positive. With this refreshed data, Numeric decided to cover 115,500 shares of its short
position before the market opened, executing the trade as the stock rose.

In the case of Stratus Computer, an analyst revised 1997 earnings from $2.90 to $3.30 after a prior
positive surprise. Although the momentum and value scores improved, Numeric decided not to
increase its existing long position, recognizing that the change wasn’t significant enough to warrant
more exposure.
 With Micro Warehouse, the company reported a positive surprise of 23 cents compared to a 20 cent
consensus. However, Numeric was cautious because the market had been expecting a letdown. Analyst
confirmation was slow, so Numeric began to partially cover its short position, reducing it from 0.6% to
0.4% of the portfolio while waiting for validation. As more analysts confirmed the earnings were solid,
Numeric exited the position completely as the stock climbed from $15.50 to $17.50.

These examples illustrate how Numeric's trading on April 29 was systematic yet adaptable. It was
driven by model scores that were updated intraday based on earnings surprises and analyst revisions,
with quick decisions made to take advantage of short-term market inefficiencies. This approach reflects
the firm's broader philosophy of blending automated insights with human judgment during significant
events.

The April 29 trades reflect the core of Numeric’s philosophy:

 Rapid model updates based on new data

 Controlled, systematic execution

 Blending quant precision with human oversight

 Tactical exploitation of earnings-related inefficiencies

5. How does the investment process at Numeric differ from that of more traditional fund managers?

Traditional Manager Numeric Investors

Fundamental analysis, stock picking Quantitative models

Interviews, site visits Data-driven scoring (value + momentum)

Discretionary judgment Systematic with minimal human override

Slower decision-making Real-time updates and intraday trading

Numeric automates most of the process. Portfolio managers focus on data validation and execution,
not subjective judgment. This makes them faster and more scalable.

6. How does Numeric Investors monitor and control transaction costs? Was Numeric's decision to close
products to new investment justified?

They break down trading costs into:

 Commissions

 Bid-offer spreads
  Market impact

 Opportunity cost (missed trades)

How they manage:

 Momentum trades are fast and time-sensitive, leading to higher costs.

 Value trades are slow and patient, resulting in lower costs.

 They size orders carefully to avoid moving the market and review execution quarterly.

Why they closed products:

 As assets under management (AUM) grew, orders became too large compared to liquidity.

 This led to higher costs, slippage, and lost alpha.

 By closing to new investments, they protected performance for existing clients, which was a
smart and disciplined decision.

7. Is Numeric truly taking advantage of market inefficiencies, or are they being rewarded for some
systematic risk? For example, by adding liquidity to the market or being long a momentum factor?

The Efficient Market Hypothesis (EMH) claims that markets reflect all available information. If this is
true, alpha should not exist.

However, Numeric profits arise from:

 Behavioral inefficiencies, such as analyst herding and delayed reactions to earnings, which
violate the semi-strong EMH.

 Slow information diffusion leads to pricing drift that can be exploited.

 Alpha comes from selecting winners or losers before the crowd.

That said, some returns also stem from:

 Exposure to known factors like momentum or value.

 Providing liquidity in value strategies.

 Using leverage in long/short strategies.

The bottom line is that it’s a combination of factors. Partly, they exploit inefficiencies (true alpha).
Partly, they receive compensation for taking systematic risks. So, they are not completely disproving
EMH. However, they are clearly finding repeatable advantages that EMH does not fully explain.