Lag-1 Autocorrelation Calculator

About the Lag-1 Autocorrelation

This calculator measures how strongly each value in a time series relates to the value immediately before it, using lag-1 autocorrelation. A positive result means the series tends to trend or stay in the same direction from one point to the next, while a negative result means it tends to alternate. It is a quick diagnostic for whether consecutive observations in a dataset are truly independent.

How It Works

You enter a series of values in their original time order. The calculator computes the mean of the series, then measures how the deviation of each value from that mean lines up with the deviation of the very next value, summed across all consecutive pairs. That sum is divided by the total sum of squared deviations across the whole series, producing a single coefficient between -1 and 1 that is then labeled as showing little, positive, or negative serial correlation based on its size.

r1 = [sum of (x_t - mean)(x_(t+1) - mean) for consecutive pairs] / [sum of (x_i - mean)^2 across all values].

Examples

Gently trending series

With the default series (12, 14, 13, 15, 17, 16, 18, 20, 19, 21), the lag-1 autocorrelation comes out to about 0.591, which the calculator labels as positive serial correlation, consistent with the overall upward drift in the values.

Zigzag alternating series

For a strongly alternating series like 10, 20, 11, 19, 12, 18, 13, 17, the lag-1 autocorrelation works out to about -0.852, reflecting how each value tends to swing in the opposite direction from the one before it.

Advantages

Gives a single, easy-to-interpret number for whether a series has short-term memory, without requiring specialized time-series software.
Flags a common violation of the independence assumption behind many standard statistical tests and confidence intervals.
Works directly from raw ordered values, with no need to specify a model or estimate additional parameters.

Common Mistakes

Applying standard confidence intervals or hypothesis tests to a dataset with meaningful lag-1 autocorrelation as though the observations were independent, which understates the true uncertainty.
Concluding a series has no serial dependence after checking only lag-1, when correlation at longer lags or seasonal patterns can still be present.
Not removing an underlying trend before computing autocorrelation, which can produce a high coefficient that reflects the trend rather than genuine short-run persistence.

Edge Cases to Watch For

At least 3 values are required, since a lag-1 relationship needs at least two consecutive pairs to compute.
If every value in the series is identical, the sum of squared deviations is zero and the coefficient is undefined, so the calculator returns an error instead of dividing by zero.
The result reflects only lag-1 relationships; longer-range or seasonal dependence, such as a value relating strongly to the one 12 periods earlier, will not show up here.
A strong overall trend in the data inflates the lag-1 coefficient even when there is no meaningful short-term momentum beyond the trend itself, so detrending first gives a cleaner read on genuine serial dependence.

Common Use Cases

Analysts checking regression or forecasting model residuals for leftover structure that the model failed to capture.
Forecasters and quantitative analysts assessing whether a financial or economic time series shows short-term momentum.
Quality engineers checking whether consecutive measurements from a monitored process are behaving independently.

Frequently asked questions

Why does lag-1 autocorrelation matter?

It tells you whether consecutive values in a time series are related to each other, which many standard statistical tests assume is NOT the case (they assume independent observations) - strong autocorrelation means each new data point carries less genuinely new information, and can distort confidence intervals or p-values calculated as though the observations were independent.