ggplot2 coord_flip() in R: Horizontal Bar and Box Plots

ggplot2 coord_flip() in R swaps the x and y axes after the geom is drawn, so a vertical bar chart becomes a horizontal one without rewriting the aesthetic mapping. It is the simplest way to rescue long category labels that overlap on the x axis.

ggplot(df, aes(x, y)) + geom_col() + coord_flip()    # vertical bars to horizontal
ggplot(df, aes(x, y)) + geom_boxplot() + coord_flip() # vertical box to horizontal
ggplot(df, aes(x, y)) + geom_point() + coord_flip()   # rotate any geom
coord_flip(xlim = c(0, 50))                           # zoom the new x axis
coord_flip(ylim = c(0, 100))                          # zoom the new y axis
coord_flip(clip = "off")                              # let labels spill outside
coord_flip(expand = FALSE)                            # remove default padding
geom_bar(orientation = "y")                           # modern alternative, no flip

What coord_flip() does in one sentence

coord_flip() is a coordinate transformation that swaps the x and y axes after every layer is drawn. Because the swap happens at the coordinate stage rather than the aesthetic stage, you keep writing aes(x = category, y = value) and the function rotates the canvas at render time. The benefit shows up immediately on bar charts with long category names: rotating the plot moves the labels onto the vertical axis where they have room to breathe.

Syntax

coord_flip() takes four arguments and zero are required.

• xlim and ylim: act on the axes after flipping, so xlim controls what used to be the y axis.
• expand: set to FALSE for bar charts that should start exactly at zero.
• clip: switch to "off" when annotations or long bars need to render beyond the panel.

Five common patterns

Pattern 1: horizontal bar chart with long labels. The classic use case. Long category names overlap on the x axis but fit comfortably on the y axis.

Pattern 2: horizontal box plot for ordered categories. Box plots benefit from flipping when category names are long or when comparing many groups vertically.

Pattern 3: horizontal lollipop chart. A lollipop combines geom_segment() and geom_point(); flipping turns it horizontal in one line.

Pattern 4: ranked bars with forcats reorder. Reordering the factor before flipping puts the longest bar at the top, the canonical layout for ranked horizontal bars.

Pattern 5: flipping vs the modern orientation argument. Both code paths produce the same chart. The orientation route keeps the stat on the correct axis, which matters for stat_summary() and tooltips.

coord_flip() vs orientation vs swapping aes

Three routes reach the same horizontal chart, but they differ on stats and labels.

Common pitfalls

Pitfall 1: factor levels appear reversed after flipping. ggplot2 draws factor levels from bottom to top on the y axis. After coord_flip(), the first level lands at the bottom of the chart, which feels backwards for ranked bars. Fix with fct_rev() or by sorting the factor with fct_reorder().

Pitfall 2: geom_smooth() lines disappear or look wrong. geom_smooth() fits its model along the x axis, so flipping after the fit can leave the smoother in the wrong orientation. Either use geom_smooth(orientation = "y") or fit the model on the swapped mapping directly with aes(x = y_var, y = x_var).

Pitfall 3: scale limits applied to the wrong axis. After flipping, the original scale_y_continuous(limits = ...) still targets the value axis, but visually it now sits on the horizontal. New readers often pass limits to scale_x_continuous() and see no effect. Pass limits inside coord_flip(xlim = ..., ylim = ...) or remember that scale functions act on the pre-flip axes.

Try it yourself

Related ggplot2 functions

• coord_polar(): bends an axis into a circle for pies and rose plots.
• coord_fixed(): enforces a fixed aspect ratio for maps and scatter plots.
• coord_cartesian(): zooms without dropping data, no flip applied.
• geom_bar(): counts rows by category; supports orientation = "y" since 3.3.0.
• scale_y_reverse(): reverses the value axis without rotating the plot.

FAQ

When should I use coord_flip() versus geom_bar(orientation = "y")?

Use coord_flip() when the plot mixes layers that may not all support orientation, or when the project pins ggplot2 below 3.3.0. Use orientation = "y" on new code with geom_bar(), geom_boxplot(), geom_density(), or geom_smooth() because the stat runs on the correct axis and tooltips stay aligned. For a one-off chart with geom_col() only, both produce identical output, so pick the route that matches the rest of your codebase.

Why does coord_flip() reverse the order of my bars?

ggplot2 prints factor levels from bottom to top on the y axis. After flipping, the first level lands at the bottom of the chart, the opposite of what most readers expect. Wrap the category variable in forcats::fct_rev() or use fct_reorder() with the value to fix the order. Sorting the underlying data frame alone is not enough because the factor levels carry the order, not the row order.

Does coord_flip() work with facets?

Yes. coord_flip() composes cleanly with facet_wrap() and facet_grid(). Every panel rotates the same way, and free scales (scales = "free_x" or "free_y") act on the post-flip axes, which is occasionally surprising. If you want only some facets to flip, build them as separate plots and stitch with patchwork::wrap_plots().

Can I set axis limits after coord_flip()?

Yes, but inside coord_flip() itself: coord_flip(xlim = c(0, 100)). Passing limits to scale_x_continuous() after coord_flip() still acts on the original x axis (the post-flip vertical), which feels inverted. The cleanest pattern is to put display ranges inside the coord call and reserve scale_*_continuous() for transformations and breaks.

External reference: ggplot2 coord_flip() documentation.