Common mistake 1 with Line chart

More than about six lines. Past that nobody can follow any of them - draw the two you are arguing about and table the rest.

Common mistake 2 with Line chart

Connecting points that are not a sequence. A line across unordered categories implies a trend that does not exist.

Common mistake 3 with Line chart

Starting the axis away from zero and then discussing the size of the change rather than its direction. A line may start away from zero; the commentary has to be about slope when it does.

Common mistake 4 with Line chart

Plotting a nominal series over a long span without saying so. Most of the slope in a twenty-year nominal chart is inflation, and the interesting variation is what remains after it.

EconStatsCraft › Line chart

Line chart

The chart for anything with an order to it. Position along a line is the most accurately read encoding there is, which is why a line beats every alternative when the question is "what happened".

What it needs

Time or order
Series

The intuition

A line chart says two things at once: here are the values, and they are connected. The connection is a claim - that the gap between two points is a continuous journey rather than a jump - and it is true for a time series and false for categories. Most of the ways a line chart misleads come from the axis rather than the line: where zero sits, whether the scale is linear, and whether the horizontal spacing matches real time.

When to use it

Anything with an order: time, age, distance, rank. If the x axis has a natural sequence, start here.

Where the estimator comes from

Each observation is a point at (t_i, y_i) and consecutive points are joined by a straight segment. That segment is INTERPOLATION: it asserts values between the observations that were never measured. On a monthly series the assertion is mild; on annual data plotted with twelve months between marks it invents eleven readings a year. Where the horizontal variable is categorical the segment asserts an ordering and a distance that do not exist, which is why a line joining "North, South, East" is a chart of nothing.

What it assumes

The x axis really is ordered and evenly spaced, or close to it. A line drawn across gaps of unequal width implies a rate of change that is not in the data.
The gaps are gaps. A line drawn through a missing period invents the value it passes through; this app breaks the line instead.
The series are measured on a consistent basis throughout. A definitional change midway - a rebased index, a revised methodology - produces a step that looks like a finding and is an artefact of the measurement.
The reader will interpret the connecting line as interpolation. Joining annual points implies the value moved smoothly between them, which is a claim the data does not make.

Statistical properties

A truncated y-axis exaggerates change and a zero-based axis can hide it. Neither is dishonest by default: for an index sitting near 100 a zero baseline wastes the whole plot, and for a count a truncated one manufactures drama. The choice must fit the quantity.
A log scale turns equal PERCENTAGE changes into equal distances, so on a log axis a straight line is constant growth. This is the right default for anything compounding over a long span, and it is the only way to compare growth of series at different levels on one chart.
Unequal time spacing plotted at equal intervals distorts the slope. A gap in the data drawn as a straight segment reads as a smooth trend across a period nobody measured.
Too many lines is the commonest failure. Beyond about eight, colours stop being distinguishable and the chart becomes a texture; this app caps the legend and labels the ends instead.
A dual axis lets you set two arbitrary scales and thereby manufacture a visual correlation between any two series. There is no honest default for the two ranges, which is why the technique is best avoided entirely.
Missing values must be shown as a break, not bridged. A line drawn straight through a missing year is a fabricated observation.

How to read the output

Read the slope, not the level. A line chart is about change; if the question is "which is bigger", bars answer it better.
Two series in different units do not belong on one axis in levels. Rebase them to 100 instead - a dual axis lets the author choose the scaling that makes their point.
A log scale turns constant growth into a straight line, so a bend becomes a change in the growth rate. On any series that compounds, plot it both ways.
The last point is the least reliable. Recent observations are routinely revised, and a line chart gives the final value the same visual weight as one settled years ago.

A worked reading

A series running from 183 to 241 plotted on an axis from 180 to 245 looks like a near-vertical climb; the same series from zero looks almost flat. Both are the same 31.6% rise. Neither axis is a lie and the reader will take a different message from each, which is why the axis choice deserves as much thought as the analysis. The defensible rule: zero-base anything where the reader will compare BAR LENGTHS or judge magnitude, and let the data fill the frame where they will read a RATE of change.

Symptom → diagnosis → what to do

If you see this	It means	Do this
A dramatic-looking rise	Check where the y-axis starts	Re-plot from zero and see whether the story survives
Two series appearing to move together	Dual axis with independently chosen ranges	Rebase both to 100 on one axis instead
A straight segment across a gap	Missing data bridged	Break the line; do not interpolate
Growth looking like it is slowing	Linear axis on a compounding series	Use a log scale; constant growth is a straight line there
A line over categories	The connection asserts an order	Use a bar chart
Fifty lines in a tangle	Beyond visual capacity	Facet, highlight a few, or label ends only

Common mistakes

More than about six lines. Past that nobody can follow any of them - draw the two you are arguing about and table the rest.
Connecting points that are not a sequence. A line across unordered categories implies a trend that does not exist.
Starting the axis away from zero and then discussing the size of the change rather than its direction. A line may start away from zero; the commentary has to be about slope when it does.
Plotting a nominal series over a long span without saying so. Most of the slope in a twenty-year nominal chart is inflation, and the interesting variation is what remains after it.

Where to read this properly

Cleveland, The Elements of Graphing Data (1994), ch. 3.
Tufte, The Visual Display of Quantitative Information (2001).
Cairo, How Charts Lie (2019), ch. 2 and 3, on axis manipulation.

All analyses

Data & variables

Each example is simulated from a known model, so you can check whether the estimator recovers the truth - and nothing here is anyone's private data.

Filter variables

Drag a variable into a slot on the right - or just click it.

Choose an example dataset above to begin.

Choose a transformation - it opens in the data editor, with a live preview.

Apply to

Multiply by

Columns ctrl-click for several

Group by ctrl-click for several

Within groups

Statistic

Bands

How many

Cut points

Keep rows where

Value

and

Fill with

Value

Also mark which rows were filled

Rules

Base period = 100

Which part

Percent at each tail

Threshold

Survey weight optional

Direction

Names column

Values column

Largest first

Delete them, not just count them

Read this column as

Tidy the text

New name optional

EconStatsCraft rigorous, plain-spoken econometrics

Econometrics that explains itself.

Regression, panel data, time series and the whole battery of tests. Every result is followed by a verdict you can defend.

Load a dataset to see it here.

Line chart

What it needs

The intuition

When to use it

Where the estimator comes from

What it assumes

Statistical properties

How to read the output

A worked reading

Symptom → diagnosis → what to do

Common mistakes

Where to read this properly

See also

Related Charts

All analyses

Charts

Maps

Explore

Regression

Time series - one series

Time series - several series

Panel data

Multivariate statistics

Marketing research

Causal

Finance & risk

Actuarial

Tests & breaks

Machine learning

Econometrics that explains itself.