Beginning to sonify the Economist’s Big Mac index

The Big Mac index is The Economist’s index of the Purchasing-Power Parity (PPP), which was devised in 1986 . Intended as a light-hearted representation of exchange rate theory, it has become one of the economic indexes, along with the Hemline Index [1] and the Men’s Underwear Index . All three indices represent a view of the underlying economic world and represent one facet for the viewer.

Time series data is a challenge in sonification and auditory displays in two ways. Firstly, the data needs to be represented in a format that the listener understands is a data point. Secondly we have to represent the passage of time for the data as we do not have a visual stimulus for the overview.

In the initial experiment, we downloaded the CSV file of the Big Mac index [7] and extracted the UK information for the last four years. These contain two data points for each year, January and July, giving us 8 data points for this experiment.

In this short paper, we describe an experiment using The Economist’s Big Mac [6] index data to explore sonifying time series data. We talk about the approaches to the data and the issues that arise from it.

We review some related works then describe the experiment. We then discuss the results, before reviewing further work and concluding.

Related Works

In [3], the author describes the way in which caricatures of sounds can be used to help the user understand the relationship that is being described. In [5], the author describes his work with the SonicFinder sonic interface for Apple.

The Foreign Affairs website used tones and pitch to provide a time series graph of the gold and silver prices, pictured against two graphs of the data to provide the data points. The sonification gives an overall view of the changes between the two indices but the time-based axis is not addressed.

Experiment

One of the issues is the minor variance between the data values. As the price may only change by 0.2 or 0.1 in the dataset, the variance has to be represented without distorting the data.

These were then put into the Chuck language, designed for music generation and audio playback, and sonified using a standard Saw Tooth Oscillating wave. This provides us with the raw overview of the change in rate.

The sound was changed from a wave to a WAV file that had previously been recorded for another project. Each new data point determined the length of time that the recording was played with a small gap in between the data points to mark the change.

Discussion

This use of a sample suggests that sampling would be a useful way of representing the data points in a pleasant and meaningful manner. Initial thoughts are to use either the clink of money for each point or an eating sound.

Sounds of eating presents an aesthetic issue as the noise is not the most pleasant to hear. A choice may be to use a more exaggerated sound, akin to the noises used in computer or arcade games, to make it useful to the listener. As the index is a light hearted view of serious data, it seems in keeping with the data intent to use samples to provide some levity without making this more important than the data.

Informal studies were undertaken to determine sounds that might be appealing to the user to replace the sound of a real person eating. The closeness of the sound to our own habits and the psychological associations make the sound unappetizing to hear in an auditory display. These associations affect our responses to words, such as moist , and actions. Our response, following our own thoughts and suggestions, is to search for a crisper sound. A common suggestion was the use of arcade games and the character sounds, especially Pac-Man.

After experimentation with the various types of noises that the character makes eating the ghosts, cherries and dots, it was decided that whilst appealing, the sound was too abstract. This is a concern for the work, as we are not relying upon the use of visual stimuli to help the user. For rigorous auditory work, even with what was intended as a joke, there must be a line between the abstract and concrete.

The sound should be concrete enough to allow for the user to associate the sound to an event. Yet it should not be hyper-real as it is potentially uncomfortable for the user. Brewster [2] discusses the design issues in visual buttons for a graphics package to aid the user mixed with earcons to help the user. Using a sound that echoes the action, in this case a burger being eaten, aids the user in creating the link with the data but like the use of a scalpel for the cropping action, the earcon needs to be abstract enough to recognize it.

By altering the sample rate, we can alter the perception for the listener of the data. We use the bite sound for the underlying Big Mac data for the United Kingdom. In the ChucK, language, the buffer uses the rates to alter the sample’s speed. We use this rate function to the show the relationships with the data. In the Big Mac index, we use the sample rate in two ways.

The sample rate might be slower if a Big Mac is more expensive to reflect the notion of eating an expensive meal slowly. Once this relationship is known, the slower eating sound is expressive.

Alternatively the sample rate might be faster which produces a potentially more compressed sound. If the exchange rate is more expensive, then the rate is faster.

Altering the sample rate is analogous to phase offset where musical phases overlap.

As this is sound, not visual, we are not limited to 2 or 3 dimensional axes. We can also use volume as the Z-axis to further enhance the listener’s experience as the data points have so little variance. As the minimum value is 3.82 and the maximum is 4.93, the total change is only 1.11 that is then reduced to 0.111 in variance.

The second issue is the change between months and years and how this might be presented to the user. In a visual graph, we might expect the axes to have ticks, points in the range at equidistant moments along the axis, at the very least and a legend that gives the viewer a context for the data point. Auditory displays do not have this option but we might use the gaps to provide short ticks to highlight major changes.

A more successful approach is the use of a wave sound point at a year change to provide the sense of the motion. This can be use subtly with a constant time signature to provide a sense of constant time. We may think of creating major and minor ticks to help the user locate themselves within the changing data and to understand the changes in time period. This would be more ideal for a larger number of points than two per year.

Future Work

An auditory icon, or earcon, may be developed for this project using theatrical and sound engineering techniques from special effects. Although this would take extra time, it does allow the display sounds to be tailored to the graph.

Usability tests need to be conducted to test the various options to define the most useful options for the user. This would require a pilot test and a wider study.

As the data set has various countries, future work should include comparing many streams of data, such as parts of Europe compared against each other or other groups, such as the super power index.

Conclusion

Auditory displays provide ways of exploring data sets and making the underlying patterns and comparisons shown the user.

There is a still work to be done in the time series space that affects the listener’s comprehension of not only the time but also the underlying data. This work is essential to seeing auditory displays understood as an analytical tool.

Using the caricature of the biting sound helps the listener understand the data and the changes. As the changes are so minor, we find that we must use other dimensions such as volume and duration of the sample, its rate.

There is a limitation in that the icon works for singular streams rather than multiple streams of data. This might be future work but it also suggests a deeper question regarding the linking of multiple data streams into a coherent structure.

As a simple experiment, we have raised and managed to answer some questions regarding the use of auditory icons to represent data relations. We have successfully used an icon to represent the UK data on the Big Mac index. We believe that it echoes the intent of the underlying data with some humour.
References

[1] van Baardwijk, Marjolein and Franses, Philip Hans, (2010), The hemline and the economy: is there any match?, No EI 2010-40, Econometric Institute Research Papers, Erasmus University Rotterdam, Erasmus School of Economics (ESE), Econometric Institute, http://EconPapers.repec.org/RePEc:ems:eureir:20147
[2] Stephen A. Brewster. 1998. Using Earcons to Improve the Usability of a Graphics Package. In Proceedings of HCI on People and Computers XIII (HCI ’98), Hilary Johnson, Laurence Nigay, and Chris Roast (Eds.). Springer-Verlag, London, UK, UK, 287-302.
[3] William W. Gaver. 1986. Auditory icons: using sound in computer interfaces. Hum.-Comput. Interact. 2, 2 (June 1986), 167-177. DOI=http://dx.doi.org/10.1207/s15327051hci0202_3
[4] Wersényi, György , Auditory Representations of a Graphical User Interface for a Better Human-Computer Interaction in Auditory Display:5954, Lecture Notes in Computer Science, Ystad, Sølvi, Aramaki, Mitsuko, Kronland-Martinet, Richard, Jensen, Kristoffer (eds), http://dx.doi.org/10.1007/978-3-642-12439-6_5
[5] William W. Gaver. 1989. The SonicFinder: an interface that uses auditory icons. Hum.-Comput. Interact. 4, 1 (March 1989), 67-94. DOI=http://dx.doi.org/10.1207/s15327051hci0401_3
[6] http://www.economist.com/content/big-mac-index, last accessed 11 October 2015
[7] http://infographics.economist.com/2015/databank/BMfile2000-Jul2015.xls, last accessed on 6 August 2015