What Is Heteroskedasticity?

In statistics, heteroskedasticity (or heteroskedasticity) happens when the standard

deviations of a predicted variable, monitored over different values of an
independent variable or as related to prior time periods, are non-constant. With
heteroskedasticity, the tell-tale sign upon visual inspection of the residual errors is
that they will tend to fan out over time, as depicted in the image below.

Heteroskedasticity often arises in two forms: conditional and unconditional.

Conditional heteroskedasticity identifies non-constant volatility related to prior
period's (e.g., daily) volatility. Unconditional heteroskedasticity refers to general
structural changes in volatility that are not related to prior period volatility.
Unconditional heteroskedasticity is used when future periods of high and low
volatility can be identified.

KEY TAKEAWAYS

 In statistics, heteroskedasticity (or heteroskedasticity) happens when the

standard errors of a variable, monitored over a specific amount of time, are
non-constant.
 With heteroskedasticity, the tell-tale sign upon visual inspection of the
residual errors is that they will tend to fan out over time, as depicted in the
image above.
 Heteroskedasticity is a violation of the assumptions for linear regression
modelling, and so it can impact the validity of econometric analysis or
financial models like CAPM.
 

While heteroskedasticity does not cause bias in the coefficient estimates, it does
make them less precise; lower precision increases the likelihood that the
coefficient estimates are further from the correct population value.

The Basics of Heteroskedasticity

In finance, conditional heteroskedasticity is often seen in the prices of stocks and
bonds. The level of volatility  of these equities cannot be predicted over any
period. Unconditional heteroskedasticity can be used when discussing variables
that have identifiable seasonal variability, such as electricity usage.

As it relates to statistics, heteroskedasticity (also spelled heteroscedasticity) refers

to the error variance, or dependence of scattering, within a minimum of one
independent variable within a particular sample. These variations can be used to
calculate the margin of error between data sets, such as expected results and actual
results, as it provides a measure of the deviation of data points from the mean
value.
For a dataset to be considered relevant, the majority of the data points must be
within a particular number of standard deviations from the mean as described by
Chebyshev’s theorem, also known as Chebyshev’s inequality. This provides
guidelines regarding the probability of a random variable differing from the mean.

Based on the number of standard deviations specified, a random variable has a

particular probability of existing within those points. For example, it may be
required that a range of two standard deviations contain at least 75% of the data
points to be considered valid. A common cause of variances outside the minimum
requirement is often attributed to issues of data quality.

The opposite of heteroscedastic is homoskedasticity. Homoskedasticity refers to a

condition in which the variance of the residual term is constant or nearly so.
Homoskedasticity is one assumption of linear regression modelling. It is needed to
ensure that the estimates are accurate, that the prediction limits for the dependent
variable are valid, and that confidence intervals and p-values for the parameters
are valid.

The Types Heteroskedasticity

Unconditional
Unconditional heteroskedasticity is predictable and can relate to variables that are
cyclical by nature. This can include higher retail sales reported during the
traditional holiday shopping period or the increase in air conditioner repair calls
during warmer months.

Changes within the variance can be tied directly to the occurrence of particular
events or predictive markers if the shifts are not traditionally seasonal. This can be
related to an increase in Smartphone sales with the release of a new model as the
activity is cyclical based on the event but not necessarily determined by the
season.

Heteroskedasticity can also relate to cases where the data approach a boundary
where the variance must necessarily be smaller because of the boundary's
restricting the range of the data.

Conditional
Conditional heteroskedasticity is not predictable by nature. There is no telltale
sign that leads analysts to believe data will become more or less scattered at any
point in time. Often, financial products are considered subject to conditional
heteroskedasticity as not all changes can be attributed to specific events or
seasonal changes.
A common application of conditional heteroskedasticity is to stock markets, where
the volatility today is strongly related to volatility yesterday. This model explains
periods of persistent high volatility and low volatility.

Special Considerations
Heteroskedasticity and Financial Modelling
Heteroskedasticity is an important concept in regression modelling, and in the
investment world, regression models are used to explain the performance of
securities and investment portfolios. The most well-known of these is the  Capital
Asset Pricing Model (CAPM), which explains the performance of a stock in terms
of its volatility relative to the market as a whole.1 Extensions of this model have
added other predictor variables such as size, momentum, quality, and style (value
versus growth).

These predictor variables have been added because they explain or account for
variance in the dependent variable. Portfolio performance is explained by CAPM.
For example, developers of the CAPM model were aware that their model failed
to explain an interesting anomaly: high-quality stocks, which were less volatile
than low-quality stocks, tended to perform better than the CAPM model predicted.
CAPM says that higher-risk stocks should outperform lower-risk stocks.

In other words, high-volatility stocks should beat lower-volatility stocks. But

high-quality stocks, which are less volatile, tended to perform better than
predicted by CAPM.

Later, other researchers extended the CAPM model (which had already been
extended to include other predictor variables such as size, style, and momentum)
to include quality as an additional predictor variable, also known as a "factor."
With this factor now included in the model, the performance anomaly of low
volatility stocks was accounted for. These models, known as multi-factor models,
form the basis of factor investing and smart beta.
 What Is Autoregressive Conditional Heteroskedasticity
(ARCH)?
Autoregressive conditional heteroskedasticity (ARCH) is a statistical model used
to analyze volatility in time series in order to forecast future volatility. In the
financial world, ARCH modelling is used to estimate risk by providing a model of
volatility that more closely resembles real markets. ARCH modelling shows that
periods of high volatility are followed by more high volatility and periods of low
volatility are followed by more low volatility.

In practice, this means that volatility or variance tends to cluster, which is useful

to investors when considering the risk of holding an asset over different time
periods. The ARCH concept was developed by economist Robert F. Engle III in
the 1980s. ARCH immediately improved financial modelling, resulting in Engle
winning the 2003 Nobel Memorial Prize in Economic Sciences .

KEY TAKEAWAYS

 Autoregressive conditional heteroskedasticity (ARCH) models measure

volatility and forecast it into the future.
 ARCH models are dynamic, meaning they respond to changes in the data.
 ARCH models are used by financial institutions to model asset risks over
different holding periods.
 There are many different types of ARCH models that alter the weightings to
provide different views of the same data set.

Understanding Autoregressive Conditional Heteroskedasticity

(ARCH)
The autoregressive conditional heteroskedasticity (ARCH) model was designed to
improve econometric models by replacing assumptions of constant volatility with
conditional volatility. Engle and others working on ARCH models recognized that
past financial data influences future data—that is the definition of autoregressive.
The conditional heteroskedasticity portion of ARCH simply refers to the
observable fact that volatility in financial markets is non-constant all financial
data, whether stock market values, oil prices, exchange rates, or GDP, go through
periods of high and low volatility. Economists have always known the amount of
volatility changes, but they often kept it constant for a given period because they
lacked a better option when modelling markets.

ARCH provided a model that economists could use instead of a constant or

average for volatility. ARCH models could also recognize and forecast beyond the
volatility clusters that are seen in the market during periods of financial crisis or
other black swan events. For example, volatility for the S&P 500 was unusually
low for an extended period during the bull market from 2003 to 2007, before
spiking to record levels during the market correction of 2008. 2  This uneven and
extreme variation is difficult for standard-deviation-based models to deal with.
ARCH models, however, are able to correct for the statistical problems that arise
from this type of pattern in the data. Moreover, ARCH models work best with
high-frequency data (hourly, daily, monthly, quarterly), so they are ideal for
financial data. As a result, ARCH models have become mainstays for modelling
financial markets that exhibit volatility (which is really all financial markets in the
long run).

The Ongoing Evolution of ARCH Models

According to Engle's Nobel lecture in 2003, he developed ARCH in response to
Milton Friedman's conjecture that it was the uncertainty about what the rate of
inflation would be rather than the actual rate of inflation that negatively impacts
an economy.3 Once the model was built, it proved to be invaluable for forecasting
all manner of volatility. ARCH has spawned many related models that are also
widely used in research and in finance, including GARCH, EGARCH, STARCH,
and others.

These variant models often introduce changes in terms of weighting and

conditionality in order to achieve more accurate forecasting ranges. For example,
EGARCH, or exponential GARCH, gives a greater weighting to negative returns
in a data series as these have been shown to create more volatility. Put another
way, volatility in a price chart increases more after a large drop than after a large
rise. Most ARCH model variants analyze past data to adjust the weightings using
a maximum likelihood approach . This results in a dynamic model that can forecast
near-term and future volatility with increasing accuracy—which is, of course, why
so many financial institutions use them.