An ordered horizontal axis changes what the marks mean. Position along that axis now encodes time, the gaps between points are meaningful, and a line drawn between two points asserts that the quantity passed through the values in between. That assertion is usually true for a stock price and usually false for a survey run twice a year. It's the first thing to check.
When do lines win, and when do columns?
Lines win for five or more points with four or fewer series. They encode change as slope, which the eye reads well, and they show the shape of a series and not just its endpoints: the plateau, the inflection, the month everything turned. No other form in the family shows any of it.
Columns beat lines when the periods are discrete events, as opposed to samples of a continuous quantity, and when there are few enough of them to read individually. Quarterly dividends are events. Monthly closing prices are samples of something that existed every day. Draw dividends as columns and prices as a line, and the form tells your reader something true about the data before they've read a single number.
Advertisement
What does a slope chart show, and what does it hide?
Slope charts handle exactly two periods. Two vertical positions, one line per category connecting them, and the reader gets rank change and magnitude change in the same picture. They're excellent at what they do. They also throw away everything in between, and that makes them dangerous when the path is the story.
Consider a market-breadth series, the share of an index's constituents trading above a long-term average, measured monthly for a year. It starts at 71 and ends at 38. A slope chart draws one straight line and implies a steady decline of 3.00 points a month for eleven months. The actual series rose to a peak of 74 in April, held above 66 until July, and then fell 28 points in five months. Eighty-five percent of the year's decline happened in the last five months, and one month alone gave up 6 points, twice the implied constant rate. Those are the facts that make the series interesting, and the two-period form deletes every one of them.
When is an area chart the wrong choice?
Area charts fill the space under a line, and that fill is a claim. Area encodes magnitude, so the reader now reads the quantity of shading as meaning something, and the chart acquires an obligation the line did not have: it must be anchored at zero. For the breadth series that's ruinous. A share needs an axis running 0 to 100, so the 36-point range that carries the entire story is compressed into 36% of the plot height, and 51.4% of the fill at its tallest sits below the series minimum, encoding nothing. Use area when the fill really does mean a cumulated quantity, most often for a part-to-whole series where the bands stack to a total. For a single percentage, the line is correct and the fill is decoration that costs you resolution.
What is the sampling question?
Underneath all of this is the sampling question. Before you choose a form, ask what interval the data was actually observed at, whether the gaps are equal, and whether the interval is fine enough to show the movement you are claiming. A chart of four annual points cannot support a claim about a turning point. A chart of daily points cannot support a claim about a trend unless you say what you smoothed and how. Neither of those is a formatting question, and both of them decide whether your chart tells the truth.