What Makes a Holiday Happy?
Whether or not you're happy this holiday season, you've probably experienced pressure to feel that way. It's inescapable. The ads, the songs, the movies.... and all those well-meaning folks who, by wishing you a "happy holiday", imply that your holiday should be happy.
Don't get me wrong – I'm generally content, and I like the holidays. But I also reserve the right to be cranky about the pandemic, the Republican party, climate change, Christmas muzak, and iced sugar cookie almond milk lattes (available now at Starbucks. Seriously.). Psychologists often note that pressure to feel happy duing the holidays exacerbates the less-than-happy feelings many of us have.
In fact, we experience pressure to be happy all year long. There's even a peer-reviewed academic journal called the Journal of Happiness Studies, where social scientists publish research on how to attain that often-discussed, somewhat elusive state of being. (Happiness Studies....no pressure there, right?)
This week's newsletter is about a study that addresses a simple question: What makes Christmas merry? Although at first I'll sound like Scrooge as I point out some fatal statistical flaws, I will argue that in the end, in spite of these flaws, the study does reveal something about what makes Christmas (and other holidays) more or less merry, and it even suggests how we might avoid the usual pressures around holiday cheer.
The Merry Christmas study
This study, entitled "What makes a merry Christmas?", was published by Knox College researchers in the Journal of Happiness Studies in 2002. (Ordinarily I focus on current studies; I'm making an exception this time for obvious reasons.)
Broadly speaking, the study methodology was sensible: 117 people ranging in age from 18 to 80 were surveyed on what they did and how they felt during the Christmas season (defined on the surveys as December 12 through 27). The surveys consisted of four parts:
—Participants described how often they engaged in activities pertaining to family, religion, tradition, spending, receiving, helping, and enjoying.
—Participants estimated how much they spent on gifts and charitable contributions, as well as the cash value of the gifts they personally received.
—Participants noted whether or not they engaged in environmentally conscious holiday consumption (such as giving eco-friendly gifts, agreeing with family to set limits on spending, using alternatives to wrapping paper, etc.).
—Participants described their feelings during the Christmas season. Based on their descriptions, the researchers created a variable they called "Christmas Well-being”, or CWB. In some places they also referred to it as "merriness", or "happiness". (Appendix 1 contains further details on measurement.)
The researchers found that people showed higher CWB when they engaged in more family activities, more religious activities, or more environmentally-conscious consumption. At the same time, people reported lower CWB when they devoted more time to buying or receiving gifts.
In short, the researchers concluded that you'll have a happier holiday if you spend more time engaged with family, religion, and/or environmentally-conscious activities, and less time focused on buying and/or receiving gifts. That's a very traditional set of conclusions…
Now I'm going to channel Scrooge for a bit and explain why the stats are deeply flawed. Then I'll show that the researchers do, inadvertently, reveal something important about what makes for a more or less happy holiday.
Weak associations
All of the correlation coefficients, without exception, were small. For example, the correlation between CWB and time spent with family was 0.19. In other words, spending more time with family was associated with greater happiness, but the extent of association was close to zero.
What does that 0.19 mean exactly? Well, by squaring a correlation coefficient, you can see how much variation in one variable is predictable from the other one. 0.19 squared is .036, which means that only 3.6% of variability across people in CWB can be attributed to time spent with family. 96.4% of that variability is determined by other things. (Those "other things" don't just include non-family variables. According to some studies, how you feel when you're with your family has a much stronger impact on your happiness than the amount of time you spend with them.)
Failure to guard against flukes
The more statistical tests you run, the more likely it is that any one test will yield a significant result, just by chance.
Here's an analogy: If I toss an ordinary coin 6 times, record the results, then do the same thing for many more sets of 6 tosses, I will eventually get heads 6 out of 6 times. It won't happen very often – the chances of 6 heads in a row are 1/64 – but it will happen if I keep tossing the coin. However, once it does happen, I shouldn’t assume that I have an unusual coin (i.e., one that’s more likely than chance to turn up heads), because I already know that the more I toss it, the more likely flukes like this will occur.
The chances of such flukes occurring can be minimized by various statistical precautions. Unfortunately, the researchers didn't pursue any of them, though they ran 16 separate correlational analyses for CWB. If they had taken standard precautions, the correlation between family time and CWB would no longer be significant! (The same holds for the correlation they reported between CWB and environmentally conscious consumption.) Appendix 2 explains these statistical "precautions" in further detail. The main point is that the researchers wouldn’t have had much to report if they had run their stats properly.
Ambiguous causality
The researchers assumed that your activities influence your mood. However, it's also clear that your mood influences how much you pursue certain activities. The stats used here don't clearly support either interpretation over the other. For example, consider again that positive correlation between family time and well-being. The researchers argued that the more time you spend with your family, the better you'll feel, but it might actually be the other way around. It might be that the better you feel, the more time you spend with your family. The same ambiguity crops up for each of the other significant correlations reported in the study. If spending more money is associated with less happiness, if could be that spending money diminishes happiness, but it could just as well be that when people are less happy, they spend more money. Perhaps it's a little of both, depending on the person. There's no way of knowing from these findings.
What does the Merry Christmas study actually show?
I'm struck by the consistently low correlations reported in this study. You might expect that in a study with weak methodology, insufficient protection against flukes, etc, at least a few analyses would reveal moderate to large coefficients. The fact that they're all small (< .30) suggests to me that the researchers were consistently looking in the wrong places for effects.
In other words, perhaps no activity is special, in the sense that you become more or less happy simply by doing more of it. Spending time with family or engaged in religious activities might make you more happy, or less happy, or have no impact at all. Presumably, the quality of the experience is more important than the quantity. Likewise, buying and receiving gifts might make you less happy, more happy, or emotionally unchanged, depending on your values, your interests, the extent to which you feel obliged, and so on. (What I’m suggesting here is sort of obvious, but also consistent with the researchers’ uniformly low correlation coefficients.)
In short, if happiness is the goal, there isn't any particular thing you should be doing more or less of during the holiday. Sure, better studies than this one suggest that being with or supporting others tends to make people feel good, but in the end your "holiday well-being" will be determined by what works best for you, not by what anyone says or implies that you should do.
And, so I hope that, to the greatest extent possible, you have the holiday you want to have. If that means trying an iced sugar cookie almond milk latte, great, I hope you enjoy it. I’ll be here muttering "humbug" under my breath!
Appendix 1: The measurement of Christmas Well-being
The nice thing about peer-reviewed research, as opposed to, say, conversations around the fireplace, is that you don't have to worry that phrases like "Christmas Well-being" sound vague. Peer-reviewed studies almost always contain operational definitions – i.e., concrete descriptions of how variables are measured. Thus, we shouldn't pick on (or praise) this study’s focus on "Christmas Well-being" until we've looked at the operational definition.
To create a Christmas Well-being (CWB) score, the researchers measured satisfaction, positive affect, negative affect, and stress. For each of these four variables, participants rated how well a set of statements described how they felt over the holiday. For example:
"I am satisfied with this Christmas season".
"I feel enthusiastic [this Christmas season]".
"I feel sad [this Christmas season]".
"I am stressed this Christmas season."
Participants rated each statement on a 5-point scale (a "1" meant that the statement doesn't reflect how they felt, while a "5" meant that the statement captures their feelings extremely well). The researchers then averaged and standardized responses for items corresponding to each variable, and came up with this:
Christmas Well-being = (Satisfaction & Positivity) – (Stress & Negativity).
One advantage of this approach is that each of the four contributors to CWB operates independently. This seems consistent with what we notice among people around us. For example, consider the distinction between positive and negative affect. Some people are utterly enthralled by Christmas (high positivity, low negativity), while others act like Scrooge (low positivity, highly negativity). Some people are delighted about holiday festivities but perpetually furious at their in-laws (high positivity, high negativity). And, some folks get so overwhelmed by holiday obligations that they end up feeling pretty numb (low positivity, low negativity).
A Christmas Carol provides further illustration. Bob Cratchit, for instance, is fairly satisfied, extremely positive, remarkably low in negativity, and quite stressed by his employer, his low wages, and Tiny Tim's future. Mrs. Cratchit is both less positive and more negative. When Bob offers a toast to Scrooge, acknowledging that his employer made their little feast possible, Mrs. Cratchit gets upset and expresses the desire to give Scrooge a piece of her mind to feast on. Bob, relentlessly positive, encourages his wife to be nice in front of the children, but she continues to malign Scrooge (and Bob continues to ask her to make nice). Scrooge, when you first meet him, is perpetually dissatisfied, low in positivity, maximally negative, and somewhat stressed by his holiday business obligations. By the end of the story, Scrooge's CWB is pretty high.
I have mixed feelings about the operational definition of CWB. On the one hand, the distinctions between the four variables described above seem useful. On the other hand, the rating system is superficial. How enthusiastic did you feel between December 12 and 27? How sad did you feel? It's hard to imagine how a single number, drawn from a 5-point scale, could adequately capture it, given that emotions like enthusiasm, sadness, joy, etc. vary in duration and intensity from day to day. The researchers would’ve been better off using more sophisticated measures, or at least obtaining a larger sample so they could focus on extremes (i.e., folks with the highest and lowest CWB scores).
Appendix 2: Failure to prevent Type 1 error
Statistically speaking, the researchers should have guarded against Type I error – in this case, reporting a significant correlation when the association between the two variables is just a fluke. One of the most widely used precautions is a mathematical correction that would, in effect, make it less likely that any one analysis would yield a significant result.
(If you're a stats person, you know that I’m referring to the Bonferroni correction, which adjusts the alpha level according to the number of analyses run. Since there were 16 correlational analyses, alpha would be .05/16 = .003. However, the p value for family time and CWB was reported as being between .01 and .05.)
In short, the family time-CWB correlation wouldn’t have been significant if a conventional correction had been used. (The same is true for the correlation between environmentally conscious consumption and CWB. As for the other significant correlations, it's impossible to tell. For each of these correlations, p was simply reported as "< .01", and so we don't know whether it was less than .003 or not. Either way, the researchers end up with a few lumps of coal in their stockings!)