Modeling Apple Stock Prices with Geometric Brownian Motion

A practical introduction to GBM and Monte Carlo simulation in quantitative finance

Can we predict the future price of a stock?

Not with certainty. But we can model a distribution of possible outcomes.

That distinction is fundamental in quantitative finance. Financial markets are inherently uncertain, and stock prices continuously respond to new information, changes in expectations, macroeconomic developments, company-specific news, and unexpected shocks.

Rather than attempting to identify one exact future price, quantitative models can help us answer a more useful question:

What range of future outcomes is consistent with the behavior of the stock and the assumptions of our model?

One of the classic models used to approach this problem is Geometric Brownian Motion (GBM).

GBM is a benchmark model for stock-price dynamics and also provides an important mathematical foundation for the Black–Scholes–Merton option-pricing framework.

In this article, we will explore the intuition and mathematics behind Geometric Brownian Motion and see how it can be combined with Monte Carlo simulation to generate possible future stock-price paths.

For the applied example, we use real historical data from Apple Inc. (AAPL) and simulate possible price trajectories over approximately one trading year.

Apple Stock as Our Applied Example

Before introducing the model, it is useful to look at the asset we are trying to simulate.

In the applied exercise, we use historical adjusted closing prices for Apple.

Historical prices provide the raw information we need to estimate two fundamental inputs of the GBM model:

expected return and volatility.

However, we do not estimate these parameters directly from the level of the stock price. Instead, we work with stock returns.

We will return to this point shortly.

The Mathematics of Geometric Brownian Motion

A stock differs fundamentally from an asset that promises predetermined cash flows.

Its future value is uncertain.

To represent this uncertainty mathematically, Geometric Brownian Motion describes the stock price using a stochastic differential equation, or SDE.

The standard GBM process is

At every period, the model generates a new random shock.

Some shocks will be positive.

Others will be negative.

The parameter σ\sigmaσ determines how strongly those shocks affect the stock price.

A stock with higher estimated volatility will generate a wider range of simulated price trajectories, while a stock with lower volatility will tend to produce smoother and more concentrated paths.

This leads to a simple interpretation of GBM:

Geometric Brownian Motion combines an expected trend with a sequence of random market shocks.

Why Do We Need Itô’s Lemma?

The stochastic differential equation gives us the dynamics of the stock price, but it is not yet the most convenient form for simulation.

Our objective is to obtain an explicit expression for the future stock price StS_tSt​.

The difficulty comes from Brownian motion.

Brownian paths are not smooth, which means we cannot simply apply ordinary calculus in the same way we would with a conventional differential equation.

This is where Itô’s Lemma becomes essential.

Itô’s Lemma can be thought of as the stochastic counterpart of the ordinary chain rule.

To solve the GBM process, we apply Itô’s Lemma to the logarithm of the stock price:

f(St)=ln⁡(St)f(S_t)=\ln(S_t)f(St​)=ln(St​)

Doing so gives us

dln⁡(St)=(μ−σ22)dt+σ dWt\boxed{ d\ln(S_t) = \left( \mu-\frac{\sigma^2}{2} \right)dt + \sigma\,dW_t }dln(St​)=(μ−2σ2​)dt+σdWt​​

An important new term appears:

−σ22-\frac{\sigma^2}{2}−2σ2​

This is known as the Itô correction.

It arises because Brownian motion has mathematical properties that differ from those of smooth deterministic processes.

After integrating the equation and transforming back from logarithms to prices, we obtain the exact solution of Geometric Brownian Motion:

St=S0exp⁡[(μ−σ22)t+σt Z]\boxed{ S_t = S_0 \exp \left[ \left( \mu-\frac{\sigma^2}{2} \right)t + \sigma\sqrt{t}\,Z \right] }St​=S0​exp[(μ−2σ2​)t+σt​Z]​

with

Z∼N(0,1)Z\sim N(0,1)Z∼N(0,1)

This equation is extremely important.

It gives us a direct way to simulate possible future stock prices.

We need four elements:

the initial stock price S0S_0S0​, the expected return μ\muμ, the volatility σ\sigmaσ, and a random shock ZZZ.

The next question therefore becomes:

Where do μ\muμ and σ\sigmaσ come from?

Estimating Drift and Volatility from Historical Data

The two key parameters of the model are estimated from historical stock returns.

Instead of using simple price changes, GBM naturally works with log returns.

The daily log return is

rt=ln⁡(StSt−1)\boxed{ r_t = \ln \left( \frac{S_t}{S_{t-1}} \right) }rt​=ln(St−1​St​​)​

Under the standard GBM framework, log returns are normally distributed.

With approximately 252 trading days per year, the model implies

rt∼N(μ−σ2/2252,σ2252)\boxed{ r_t \sim N \left( \frac{\mu-\sigma^2/2}{252}, \frac{\sigma^2}{252} \right) }rt​∼N(252μ−σ2/2​,252σ2​)​

This relationship allows us to estimate the parameters from historical data.

Estimating Volatility

The historical standard deviation of daily log returns gives us a daily measure of volatility.

To annualize it, we multiply by the square root of 252:

σ^=sd(rt)252\boxed{ \hat{\sigma} = sd(r_t)\sqrt{252} }σ^=sd(rt​)252​​

This is the annualized volatility estimate used in the simulation.

Estimating the Drift

The average daily log return does not estimate μ\muμ directly.

Instead, because of the Itô correction, it estimates

μ−σ22\mu-\frac{\sigma^2}{2}μ−2σ2​

Therefore, the GBM drift can be recovered using

μ^=252 r‾+σ^22\boxed{ \hat{\mu} = 252\,\overline{r} + \frac{\hat{\sigma}^{2}}{2} }μ^​=252r+2σ^2​​

This distinction is important.

The mean log return and the GBM drift are related, but they are not identical.

An Important Limitation: Historical Parameters Are Not Future Parameters

At this point, we have estimated the two main parameters required by the model.

But there is an important qualification.

Both μ\muμ and σ\sigmaσ are estimated from historical data.

They describe the behavior of the stock during the selected estimation period.

There is no guarantee that the same expected return or volatility will prevail in the future.

This makes GBM simulations highly sensitive to the calibration window.

Suppose, for example, that volatility is estimated during a period of extreme market turbulence.

The estimated value of σ\sigmaσ may be relatively large, producing very wide simulated distributions.

If instead the estimation window contains a period of unusual stability, the resulting simulations may produce much narrower uncertainty ranges.

The selection of the estimation period therefore matters considerably.

This is one reason why backtesting, sensitivity analysis, and alternative calibration windows are valuable when working with these models.

From Geometric Brownian Motion to Monte Carlo Simulation

Once the current stock price and the parameters μ\muμ and σ\sigmaσ have been estimated, we can move to the simulation stage.

Instead of asking:

What will Apple stock be worth one year from now?

we ask:

What is the distribution of possible Apple stock prices one year from now?

This is precisely the type of question that Monte Carlo simulation is designed to address.

For each trading day, we simulate the next stock price using

St+Δt=Stexp⁡[(μ−σ22)Δt+σΔt Zt]\boxed{ S_{t+\Delta t} = S_t \exp \left[ \left( \mu-\frac{\sigma^2}{2} \right)\Delta t + \sigma\sqrt{\Delta t}\,Z_t \right] }St+Δt​=St​exp[(μ−2σ2​)Δt+σΔt​Zt​]​

where

Zt∼N(0,1)Z_t\sim N(0,1)Zt​∼N(0,1)

A new random shock is generated at every step.

One simulated sequence of shocks generates one possible future stock-price path.

If we repeat the process many times, we generate many alternative futures.

500 Possible Futures for Apple Stock

In the applied exercise, I generate 500 possible future Apple stock-price paths over approximately 252 trading days.

Every simulated path starts from the same current stock price.

After that point, however, each path evolves differently because it receives a different sequence of random shocks.

[INSERT IMAGE: 500 MONTE CARLO PATHS]

Figure 2. Monte Carlo simulation of 500 possible future Apple stock-price paths using Geometric Brownian Motion.

This graph illustrates one of the central ideas behind probabilistic forecasting.

There is not one future.

There is a distribution of possible futures.

Some simulated paths produce strong stock-price appreciation.

Others generate relatively flat outcomes.

Some produce substantial declines.

The objective is not to argue that any one individual path will occur.

Instead, the collection of paths provides information about the distribution implied by the model.

Summarizing Hundreds of Simulated Paths

Although plotting hundreds of simulated trajectories is informative, it can quickly become visually overwhelming.

A better approach is to summarize the distribution using percentiles.

For each future trading day, we can calculate statistics such as:

• the 5th percentile

• the 25th percentile

• the median

• the 75th percentile

• the 95th percentile

The median provides a convenient measure of the central simulated outcome.

The 25th–75th percentile interval represents a relatively concentrated range around the center of the distribution.

The 5th–95th percentile interval provides a much wider measure of uncertainty.

This produces a forecast chart that is considerably easier to interpret than hundreds of individual paths.

Connecting the Forecast with the Actual Apple Stock Price

The final visualization combines Apple's observed historical price series with the out-of-sample GBM simulation.

[INSERT IMAGE: HISTORICAL PRICE + GBM FORECAST]

Figure 3. Apple historical stock price together with the GBM median forecast and selected uncertainty scenarios.

The vertical separation between the historical and simulated sections helps distinguish observed data from model-generated outcomes.

The central path represents the median of the Monte Carlo distribution, while the surrounding ranges illustrate uncertainty.

In my applied example, the estimated distribution contains substantial upside as well as meaningful downside scenarios.

That is precisely the point.

A stochastic model should not only tell us what might happen under favorable conditions.

It should also show us what could happen when shocks move in the opposite direction.

Is the Median a Stock Price Prediction?

Not exactly.

This is an important distinction.

The median simulated path is a useful summary statistic, but it should not be interpreted as a guaranteed future price.

If the median simulation suggests that Apple could trade at a certain value one year from now, this does not mean that the model is claiming:

“Apple will trade at exactly this price.”

Instead, the correct interpretation is:

Given the estimated historical drift and volatility, together with the assumptions of GBM, this value represents the median of the simulated distribution.

This distinction between a deterministic prediction and a probabilistic simulation is essential in quantitative finance.

What Does Geometric Brownian Motion Assume?

The elegance of GBM comes partly from its simplicity.

But that simplicity also creates limitations.

Understanding these assumptions is just as important as understanding the model itself.

Constant Drift

The model assumes that the expected return parameter μ\muμ remains constant over the forecasting horizon.

In real financial markets, expected returns may change over time.

Constant Volatility

GBM assumes that σ\sigmaσ is constant.

Real financial markets frequently exhibit volatility clustering, where periods of high volatility tend to be followed by other periods of high volatility.

Models such as GARCH are specifically designed to capture this behavior.

Normally Distributed Log Returns

The standard GBM framework implies normally distributed log returns.

Actual financial returns frequently display fat tails, meaning extreme events occur more frequently than a normal distribution would suggest.

Continuous Price Paths

Perhaps one of the most important assumptions is that GBM generates continuous price paths.

Real stock prices can experience sudden jumps.

For example, an earnings announcement, regulatory decision, merger announcement, geopolitical shock, or major macroeconomic event can cause a stock price to move abruptly.

Standard GBM does not explicitly model these jumps.

This limitation motivates an important extension of the model: the Merton Jump-Diffusion Model, which introduces discrete Poisson jumps into the stock-price process.

Why Is GBM Still So Important?

Given these limitations, why do quantitative finance courses continue to teach Geometric Brownian Motion?

Because it provides an exceptionally useful benchmark.

GBM introduces several concepts that become fundamental throughout quantitative finance:

• stochastic processes

• Brownian motion

• drift and volatility

• Itô’s Lemma

• stochastic differential equations

• Monte Carlo simulation

• probabilistic forecasting

• option-pricing foundations

Once these concepts are understood, it becomes much easier to move toward more sophisticated models.

In this sense, GBM is not necessarily the final model.

It is the starting point.

From Theory to Real Financial Data

This is also why I believe quantitative finance is best learned by combining theory and application.

Understanding the equation

dSt=μStdt+σStdWtdS_t=\mu S_tdt+\sigma S_tdW_tdSt​=μSt​dt+σSt​dWt​

is important.

But it becomes much more intuitive once we estimate the parameters from an actual stock, generate hundreds of possible future trajectories, and observe how changes in volatility affect the width of the resulting distribution.

In the complete tutorial, I implement the model using R and real Apple stock data.

The workflow includes:

1. Importing historical financial data

2. Calculating daily log returns

3. Estimating annualized volatility

4. Estimating the GBM drift

5. Generating random market shocks

6. Simulating 500 future stock-price trajectories

7. Calculating forecast percentiles

8. Constructing uncertainty bands

9. Combining the historical stock price and the simulated forecast in a single visualization

The R implementation is intentionally not reproduced in this article.

The objective here is to provide the theoretical framework and intuition behind the model.

Watch the Complete Applied Tutorial

If you would like to see how the entire process is implemented using real Apple stock data in R, you can watch the complete tutorial on my Forecasting Economics YouTube channel.

In the video, we estimate the GBM parameters directly from market data, generate 500 Monte Carlo simulations, and examine the range of possible Apple stock prices over the next year.

🎥 Watch the full tutorial:

[INSERT YOUTUBE LINK]

Get the Complete R Code and Presentation

If you would like to reproduce the analysis yourself, the complete material used in the tutorial is also available.

The package includes:

✓ Complete R script

✓ Full presentation slides

✓ Parameter estimation

✓ Monte Carlo simulation

✓ 500 simulated stock-price paths

✓ Median and uncertainty-band calculations

✓ Historical price and forecast visualization

📚 Get the complete tutorial material here:

[INSERT PURCHASE LINK]

Purchasing the material also helps support Forecasting Economics and allows me to continue producing applied tutorials in econometrics, forecasting, and quantitative finance.

Final Thoughts

We cannot know Apple's future stock price with certainty.

But quantitative finance gives us tools for approaching uncertainty systematically.

Geometric Brownian Motion provides a simple but powerful framework for combining an expected return with random market shocks.

Monte Carlo simulation then allows us to take that stochastic model and generate hundreds—or thousands—of possible future outcomes.

The result is not a guaranteed prediction.

It is something arguably more useful:

a distribution of possible futures.

And that probabilistic way of thinking is at the heart of quantitative finance.