How to write a statistical story
Do you have a story?
First and foremost, you need a story to tell. You should think in terms of issues or themes, rather than a description of data. Specifically, you need to find meaning in the statistics. A technical report is not a story, nor is there a story in conducting a survey.
A story tells the reader briefly what you found and why it is important to the reader. Focus on how the findings affect people. If readers are able to relate the information to important events in their life, your article becomes a lot more interesting.
Statistical offices have an obligation to make the data they collect useful to the public. Stories get people interested in statistical information and help them to understand what the information means in their lives. After they read good statistical stories, people should feel wiser and informed, not confused.
Possible topics/themes for stories:
- Current interest (policy agenda, media coverage, etc.);
- Reference to everyday life (food prices, health, etc.);
- Reference to a particular group (teens, women, the elderly, etc.);
- Personal experiences (transportation, education, etc.);
- Holidays (Independence Day, Christmas, etc.);
- Current events (statistics on a topic frequently in the news);
- Calendar themes (spring, summer, etc.);
- New findings;
- A regular series (“This is the way we live now”, “Spotlight on xxxx”, etc.).
Write like a journalist: The “inverted pyramid”
How can statisticians communicate like journalists? By writing their stories the way journalists do. The bonus is that the media are more likely to use the information.
Journalists use the “inverted pyramid” style. Simply, you write about your conclusions at the top of the news story, and follow with secondary points in order of decreasing importance throughout the text.
Think of a typical analytical article as a right-side-up pyramid. In the opening section, you introduce the thesis you want to prove. In following sections, you introduce the dataset, you do your analysis and you wrap things up with a set of conclusions.
Journalists invert this style. They want the main findings from those conclusions right up top in your news story. They don’t want to have to dig for the story.
You build on your story line throughout the rest of the text. If the text is long, use subheadings to strengthen the organization and break it into manageable, meaningful sections. Use a verb in subheadings, such as: “Gender gap narrows slightly.”
The lead: Your first paragraph
The first paragraph, or lead, is the most important element of the story. The lead not only has to grab the reader’s attention and draw him or her into the story, but it also has to capture the general message of the data.
The lead is not an introduction to the story. On the contrary, it should tell a story about the data. It summarizes the story line concisely, clearly and simply.
It should contain few numbers. In fact, try writing the first sentence of the lead using no figures at all.
Don’t try to summarize your whole report. Rather, provide the most important and interesting facts. And don’t pack it with assumptions, explanations of methodology or information on how you collected the data.
The lead paragraph should also place your findings in context, which makes them more interesting. Research shows that it is easier to remember a news report if it establishes relevance, or attempts to explain a particular finding. Exercise caution, though. It is not a good idea to speculate, especially if your statistical office cannot empirically establish causality, or does not produce projections.
Give enough information so the reader can decide whether to continue reading. But keep it tight. Some authors suggest five lines or fewer – not five sentences – for the opening paragraph.
Poor: A new study probes the relationship between parental education and income and participation in post-secondary education from 1993 to 2001.
Good: Despite mounting financial challenges during the 1990s, young people from moderate and low-income families were no less likely to attend university in 2001 than they were in 1993, according to a new study.
Finally: there is no contradiction between getting attention and being accurate.
Remember:
- Focus on one or two findings;
- Write in everyday language (the “popular science” level);
- Create images for your readers;
- Focus on the things you want readers to remember;
- Choose the points you think are newsworthy and timely.
Good writing techniques
Write clearly and simply, using language and a style that the layperson can understand. Pretend you are explaining your findings to a friend or relative who is unfamiliar with the subject or statistics in general. Your readers may not be expert users who often go straight to the data tables.
Terms meaningful to an economist may be foreign to a layperson, so avoid jargon. Use everyday language as much as possible. If you have to use difficult terms or acronyms, you should explain them the first time they are used.
Remember: on the Internet, people want the story quickly. Write for the busy, time-sensitive reader. Avoid long, complex sentences. Keep them short and to the point. Paragraphs should contain no more than three sentences.
Paragraphs should start with a theme sentence that contains no numbers.
Example: Norway’s population had a higher growth last year than the year before. The increase amounted to 33,000 people, or a growth rate of 0.7%.
Large numbers are difficult to grasp. Use the words millions, billions or trillions. Instead of 3,657,218, write “about 3.7 million.” You can also make data simpler and more comprehensible by using rates, such as per capita or per square mile. Some suggestions follow.
Use:
- Language that people understand;
- Short sentences, short paragraphs;
- One main idea per paragraph;
- Subheadings to guide the reader’s eye;
- Simple language: “Get,” not “acquire.” “About,” not “approximately.” “Same,” not “identical”;
- Bulleted lists for easy scanning;
- A good editor. Go beyond Spell-Check; ask a colleague to read your article;
- Active voice. “We found that…” Not: “It was found that....”;
- Numbers in a consistent fashion: For example, choose 20 or twenty, and stick with your choice;
- Rounded numbers (both long decimals and big numbers);
- Embedded quotes (these are sentences that generally explain “how” or “why”, and which journalists like to use verbatim in their news stories in quotes);
- URLs, or electronic links, to provide your reader with a full report containing further information.
Avoid:
- “Elevator statistics”: This went up, this went down, this went up;
- Jargon and technical terms;
- Acronyms;
- All capital letters and all italics: Mixed upper and lower case is easier to read;
- “Table reading”, that is, describing every cell of a complex table in your text.
Not Good: From January to August, the total square metres of utility floor space building starts rose by 20.5% from the January to August period last year.
Better: In the first eight months of 2004, the amount of utility floor space started was about 20% higher than in the same period of 2003.
Headlines: Make them compelling
If your agency’s particular style calls for a headline on top of a statistical story, here are some suggestions to keep in mind.
Readers are most likely to read the headline before deciding to read the full story. Therefore, it should capture their attention. The headline should be short and make people want to read on. It should say something about the findings presented in the article, not just the theme.
Write the headline after you have written your story. Headlines are so important that most newspapers employ copy editors who craft the headlines for every story. Because the information is likely new to them, these editors can focus more readily on the most interesting aspects of the story.
In the same vein, statistical agencies might consider a similar arrangement. The individual who writes the headline could be different than the story’s author.
Headlines should:
- Be informative, appealing, magnetic, interesting and newsy, and incorporate:
- the highest since, the lowest since…;
- something new;
- the first time, a record, a continuing trend;
- Make you want to read the story, not scare you off;
- Summarize the most important finding;
- Be no longer than one line of type;
- Not try to tell everything;
- Contain few numbers, if any at all ;
- Have a verb or implied verb.
Not Good: New report released today (the report is not the news)
Energy conservation measures widespread (too vague)
Prices up in domestic and import markets (what prices?)
Good: Gasoline prices hit 10-year low
Crime down third year in a row
July oil prices levelled off in August
Tips for writing for the Internet
The principles of good writing also apply to writing for the Internet, but keep in mind some additional suggestions.
People scan material on the Internet. They are usually in a hurry. Grabbing their attention and making the story easy to read are very important.
You also have different space limitations on the Internet than on paper. Stories that make the reader scroll through too many pages are not effective. Avoid making the reader scroll horizontally.
Format the page so the story can be printed properly, without text being cut off by margin settings. A common solution is to include a link to a ‘print friendly version’, usually another page with navigation menus and banners removed.
Write your text so the reader can get your point without having to force themselves to concentrate. Use structural features such as bulleted lists, introductory summaries and clear titles that can stand alone.
Don’t use ALL CAPITAL LETTERS on the Internet. It looks like you’re shouting. Underline only words that are electronic links. Use boldface rather than underlining for emphasis. Avoid italic typefaces because they are much harder to read.
Make sure your story is printed on a contrasting background colour: either light lettering on a dark background or the reverse. High contrast improves readability on the Internet. Make sure items are clearly dated so readers can determine if the story is current.
Graphs
A picture is indeed worth a thousand words, or a thousand data points. Graphs (or charts) can be extremely effective in expressing key results, or illustrating a presentation.
An effective graph has a clear, visual message, with an analytical heading. If a graph tries to do too much, it becomes a puzzle that requires too much work to decipher. In the worst case, it becomes just plain misleading.
Go the extra mile for your audience so that they can easily understand your point.
Good statistical graphics:
- Show the big picture by presenting many data points;
- Are “paragraphs” of data that convey one finding or a single concept;
- Highlight the data by avoiding extra information and distractions, sometimes called “non-data ink” and “chart-junk”;
- Present logical visual patterns.
When creating graphics, let the data determine the type of graph. For example, use a line graph for data over time, or a bar graph for categorical data. To ensure you are not loading too many things into a graph, write a topic sentence for the graph.
Achieve clarity in your graphics by:
- Using solids rather than patterns for line styles and fills;
- Avoiding data point markers on line graphs;
- Using data values on a graph only if they don’t interfere with the reader’s ability to see the big picture;
- Starting the Y axis scale at zero;
- Using only one unit of measurement per graphic;
- Using two-dimensional designs for two-dimensional data;
- Making all text on the graph easy to understand;
- Not using abbreviations;
- Avoiding acronyms;
- Writing labels from left to right;
- Using proper grammar;
- Avoiding legends except on maps.
For example:
Adoptions fall by 2.4% in 2003
Graph from United Kingdom Office of National Statistics. Available online at http://www.statistics.gov.uk/cci/nugget.asp?ID=592 [accessed 28 September 2005].
Tables
Good tables complement text. They should present numbers in a concise, well-organized fashion to support the analysis. Tables help minimize numbers in the statistical story. They also eliminate the need to discuss insignificant variables that are not essential to the story line.
Make it easy for readers to find and understand numbers in your table. Standard presentation tables are generally small. One decimal place will be adequate for most data. In specific cases, however, two or more decimal places may be required to illustrate subtle differences in a distribution.
Presentation tables rank data by order or other hierarchies to make the numbers easily digestible. They also show the figures that are highest and the lowest, as well as other outliers. Save large complex tables for supporting material.
Always right-justify the numbers to emphasize their architecture. The guidelines listed for graphics above, such as highlighting data by avoiding “non-data ink”, also apply to the presentation of tables.
While graphics should be accompanied by an analytical heading, titles are preferred for tables. They should be short and describe the table’s precise topic or message.
For example:
Race of Juvenile Offenders
Race of juvenile offender(s)
|
Average annual percent of violent crimes committed by juvenile(s)
% |
| . |
. |
Total
|
100.0
|
White
|
59.1
|
Black
|
25.2
|
Other
|
11.4
|
More than 1 racial group
|
2.6
|
Unknown
|
1.7
|
Table from Juvenile Victimization and Offending, 1993-2003, Bureau of Justice Statistics, Special Report, August 2005, NCJ 209468 (page 8). Available online at http://www.ojp.usdoj.gov/bjs/pub/pdf/jvo03.pdf [accessed 28 September]
Maps
Maps can be used to illustrate differences or similarities across geographical areas. Local or regional patterns, which may be hidden within tables or charts, are often made clear by using a well designed map.
Maps are a rapidly expanding area of data presentation, with methods of geographic analysis and presentation becoming more accessible and easier to use. The cost of Geographic Information Systems (GIS), or software capable of mapping statistics, has decreased rapidly in the last ten years. Mapping that was once expensive, or required specialist hardware, is now within reach of most organizations. GIS analysis and presentation are now taught in schools and universities.
Producing statistical maps can be a simple process. The most common type of statistical map is the choropleth map, where different shades of a colour are used to show contrast between regions (usually a darker colour means a larger statistical value). This type of map is best used for ratio data (e.g. population density), where the denominator is usually area (e.g. square kilometres) or population. 'Count ' data which has no denominator (e.g. total number of sheep in each region), are best illustrated using proportional or graduated symbol maps. With proportional symbol maps, the size of a symbol, such as a circle, increases in proportion to the value of the statistic. All mapping software should be capable of producing these two map types. Other types of map are possible but are best retained for specialist audiences.
When designing a map, always think about the audience and try to make it quick and easy for them to understand. If there is a natural association between a colour and a topic (e.g. blue for cold temperatures) then it would be sensible to use that colour for the legend. When choosing your legend classes, do not use complex methods unless your audience will understand them. Choosing classes of equal size, or classes containing similar numbers of events, are the most common methods. When choosing how many coloured classes to use, less is often more. Fewer classes emphasize similarity between areas and more classes emphasize the differences.
It should be possible for any statistical map to be read by a user without reference to other information and knowledge. Maps should always have a title and a legend that adequately explain the statistical units, the date that the statistical information was collected or produced and the geographic area type used. The source of statistical data should also be clearly stated. Footnotes may be used to clarify this information where needed and help to simplify titles.
Average Annual Rainfall 1961 - 1990, Europe
Graph from United Nations Economic Commission for Europe. Available online at http://www.unece.org/stats/trends2005/environment.htm [accessed 30 September 2005].
How to encourage good writing
Each statistical agency may have its own ideas on ways to reward quality writing. But here are some general suggestions.
- Set goals, such as a number of stories to be written each year
- Reward good writers for the best headline, most contributions, etc.
- Make writing an expected part of the job rather than a sideline
- Explore techniques for building enthusiasm for writing
- Show staff the results of their writing, for example, by posting in the office the newspaper or magazine coverage their stories initiated
- Provide training
Next: Writing about data
This Chapter Whole Publication (26 pages - 500kb)