One data project is the tracking of popularity of given names for newborn babies over the decades. I found this site that graphs the popularity of names over the decades since the last 1800s. This site presents a simple graphing tool that shows popularity for a large number of different names. The initial chart shows all the name in stacking area charts that show all of the names over time and hovering over the individual lines shows the name and provides a link to drill in on just that name. Overall, a nice way to explore the data.
I was drawn to this project because of a genealogical observation that I had a distant relative named Job. A couple relatives, in fact. Thinking of the biblical Book of Job, I thought it was a particularly unexpected name to give a child so I decided to look into what baby name sites had to say with it. Job has not been in the top 1000 names since the late 1800s. And yet, apparently it was relatively popular a century earlier.
In my dusty recollection of social history, I recall that it was once very popular to choose names from Biblical references. That is still present today with some popular names having links to biblical references. But earlier it was more common to use more obscure names in the Bible. The fact that the name appeared anywhere in the Bible made it a candidate for naming a new born. Again, this is just my recollection of old tales that may simply have been rumors. I haven’t done any research to substantiate any of this. For now, I’m just assuming it was true.
Even given the preference for biblical names, Job seems to be an odd name to choose. Job is hardly an obscure name with an entire book dedicated to his perseverance through hardships that fell upon his earlier prosperous life. One theory about the origin of the book of Job was that a treasured work of poetry about perseverance was preserved by giving it an additional celestial narrative to explain how Job got into his misfortunes and how he was later rewarded for his perseverance. This addition successfully preserved the work but the origin suggests three different works: a core story of a man maintaining his dignity despite a long and continuous degradation in fortunes, and intermediate version to add that this was deliberately imposed on him, and a later version that rewarded him at the end.
I have a hard time seeing how either the three stories would encourage someone to find this as an attractive name to bestow on a newborn especially with the intention of linking the baby’s life to a biblical story. I suppose Job stood out because he was a man who early on got the attention of God. A name that would draw God’s attention might be attractive in general, but it not in the sense His attention was drawn to Job.
And yet at some point, people thought it was a good idea to name their sons Job. I doubt any of my ancestors so named during the 18th century had any wealth at all. I imagine they probably lived difficult but tolerable lives. Perhaps the parents of the first Job in the family were hoping to inspire the child to persevere through the hardships they knew would be coming. That’s my hypothesis.
A generalization of this hypothesis is that a given name tells us something about the times the parents were going through. A more general hypothesis may be that the popularity of a given name may tell us something about the era as a whole. That is what drew my attention to the above site plotting name frequencies over time. The growing and then declining popularity of particular names may say something about the times when the name was given to a baby.
Just playing around with the chart a little bit, I found Lucille to be a particular clear example of the basic trend I see with many names. The name starts with relative obscurity in the late 1800s. Although the chart doesn’t show earlier times, it seems reasonable that this name arose out of obscurity sometime in the mid 1800s. It rose in popularity with increasing speed until it reached its peak around 1910s and then started its decline. The shape of the decline mirrors the shape of the increase although the decline appears faster. This is an interesting case because the name appears to have completely vanished in the 1980s as if it were a name that must be avoided. But the name is experiencing a revival lately, suggesting that the same shape may repeat in the coming decades.
Other names follow similar patterns at least to the point where a noticeable peak appears with a sharp increase and decline. It is interesting to consider what could explain the popularity curve.
First of all, it is curious to me that names would have these trajectories at all. It is conceivable that the mix of popular names that worked in any particular decade would work for all decades. Despite the number of distinct names, the various names invoke a relatively small number of qualities of beauty, intelligence, strength, gracefulness, etc. I suspect some cultures do exactly this, reusing the same basic set of names for countless generations. We don’t do that here.
The proper way to study the rise and fall of popularity of names relative their times would be study all available historical materials relevant to the different points in popularity to come up with some explanation. I don’t doubt that such studies do exist. My point here is not to try to replicate this study. Instead, I want to use this extremely limited but rich data set as an illustration of mining data.
Allow me to assume that this data set of popularity of baby names over the decades is the only data available. There is no other data that makes sense to relate to this data, to provide more dimensions of the data. This is a common scenario for real world big data sets where the actual measurements are limited. This frequency of given baby names of the decades is similar to scenarios where populations in different categories change in different time periods. We task ourselves to come up with some insight from these patterns. Often the celebrated insights or discoveries in big data analysis are not explicitly recorded information in the data but instead patterns observed about the data.
Although my illustration above picked on one name Lucille, I chose this because it illustrates basic patterns I see with a large number of names. It suggests the following explanation. Initially, there is an intention to find an uncommon name to make a baby’s name distinctive from its peers and immediate ancestors. Following that, new parents may find either the name or the young children bearing the name to be attractive for them to use for their babies. At some point, the oldest of these names enter adulthood and rise to some form of fame that encourages an explosive popularity of the name. In just a few years, the name becomes embarrassing popular as same-named children are in the same classes or other groups.
I’m at a loss to explain why the name would continue to decline to the point of vanishing. With such a large population of these names eventually entering adulthood, there must be some name-bearers who will be famous and worthy of copying their name as hopeful charm that the child will achieve a similar fame. Also, I would expect people to be inclined to honor close relatives or by reusing their names. I can see that the name necessarily declines from its peak popularity, but I would expect the name to stabilize at some moderate level of popularity. Instead, frequently the name disappears entirely and it does so with great expediency: the decline in popularity is faster than its rise.
As a person who delighted in the studies of cyclical patterns, I am inclined to imagine a cyclical pattern even though I can not explain the underlying cause. Even though I don’t understand why the name ceased to be popular, it appears to be entering a new phase of popularity. A grossly oversimplified model may suggest we will soon see a similar rapid rise in popularity of this name in the coming couple decades. This is strictly a time-series model. It doesn’t attempt to explain why. It only predicts that a cycle will repeat.
Often a time-series based interpretation is sufficient. We do this when we observed in the 1960s that city streets experience what appears to be an hour of congestion in the morning and in the afternoon, a phenomena we called rush hour. We could attempt to explain it as commuting to and from work but that explanation wasn’t necessary for planning. We only needed to know the time pattern. We still use the term rush hour even though the period is much longer than an hour, and there appear to be more than just two (let alone one) such periods. We can continue transportation planning on the time analysis of traffic flows. We are motivated to explain why the traffic is there for the different project of making policies in an attempt to influence the traffic. Based on traffic numbers alone, we largely speculate why the traffic is occurring when it is occurring.
An analogous speculation can be illustrated with the baby names. The recent rise in popularity of this one name suggests there is something very similar happening today as what happened in around the 1870s. That similarity is likely just with respect to the exotic nature of this particular name. It is now sufficiently rare as to be distinctive or it references prior popular names. The above data does not extend to the 1870s but it is easy to imagine that it might have seen a similar rise out of obscurity.
Alternatively, the speculation may be more bold to suggest that sometime about the culture of today is like the culture of the 1870s in terms of the number of people feeling now it is time to evoke the ideals behind this particular name. The modern times is like the 1870s in its desire for someone who brings inspiration, or someone who evokes the age of the Roman Empire or French influence.
I mention this alternative because it seems absurd to conjecture about a shared property of different time periods based on the popularity of a particular name (or a set of names). This is what I did at the start of the post when I contemplated why a period would justify the use of the name of Job for a child. The term absurd may be too judgmental for describing this kind of conjecture. A better term would be unjustified or unsupported.
This type of unjustified and supported claims frequently come out of big data analysis. We see a pattern of categories and then we invent some story so appealing that we can see a good fit with the data. For example, my mind is currently racing through possible stories to relate today’s cultures with the culture of the 1870s.
In the more real-world example of the observed traffic patterns, we come up with explanations such as staggered and flexible work hours, part time work, traveling consultants or salesmen, errands, or simply busier schedules with more activities scattered over town. There may be some supporting evidence for this, but the strongest evidence is the original weak evidence of the time-stamped observations of traffic flows.
Although the above descriptions suggest thus such speculations are bad (I used the term absurd), this is what makes big data analysis valuable. In earlier posts, I described this process as hypothesis discovery. A discovered hypothesis is an invented story that provides a compelling explanation for the distinctive patterns seen in available data. I like to describe such discovered hypothesis as absurd only to emphasize that the next necessary step is to test the hypothesis. I can use the word absurd because the story is already convincing on its own.
In the above examples, I could test the Job hypothesis by doing more research about the living conditions of my Job-named ancestors and their parents. I could test the similarity of the 2010s with the 1870s by studying the basic cultural themes especially among those who were choosing a particular baby name. I could test the traffic hypotheses by gathering more information about people are on roads at different times. The data analysis provided the patterns that inspired the stories that proposed hypotheses to test.
Exploring data even as simple as given-name frequencies can expose recurring patterns that could suggest explanations. Particularly good explanations become discovered hypotheses that needs testing. Also as in this example where I am pursuing a test of the hypothesis, I am satisfied with just the discovered hypothesis. I have no motivation to pursue it or use this information. Unfortunately, often we will act on the discovered hypothesis because it is so compelling. We should do so only after accepting the risk that we may embark on something that will later prove to be absurd. The discovered hypothesis should be just a story that motivates further testing and corroboration.
3 thoughts on “Reflecting on baby name trends”
Pingback: Occam’s Razor in age of big data | kenneumeister
Pingback: Reflecting on baby name trends | Hypothesis Discovery
Pingback: Occam’s Razor in age of big data | Hypothesis Discovery