# Are people voting less than they used to on Physics SE?

It seems to me that the level of voting on Physics SE is noticeably lower than on most other SE sites that I frequent and lower that it has been in the past (I have been a user on the site for several years). A casual look down the list of most recent questions shows that the vast majority are on only 0 or 1 vote. I remember a time when a question would get 1 or 2 up-votes, simply if it was constructive, valid and well-presented. In fact, I get the impression that, these days, the voting system is being used almost exclusively as a tool to 'punish' questions that are deemed unworthy.

If people aren't using the voting system, there are two potential issues I can see: firstly, there is a lack of reward/incentive for people to stick around and spend their time answering questions. Secondly, it will lead to a lack of differentiation between good/bad questions and answers - the good will get lost amidst the bad/mediocre.

So, what has happened with the voting? If people aren't using it, then it may as well not be there ...

• it also seems that not many answers get officially selected. The questioners are just too happy to see some answers and then move on. – wcc Jul 27 '18 at 21:30
• Hmm, it would be interesting if somebody who knows how to use the stat thing can get hard numbers for this. – knzhou Jul 27 '18 at 23:28
• People don't realize that there's a gold badge for voting. – Rob Jul 27 '18 at 23:52
• I can't speak for others, but I downvote bad questions, bad answers and answers to bad questions a lot. It could be driving some of the 0 or 1 score questions – Kyle Kanos Jul 28 '18 at 10:50
• @KyleKanos you don't think it's possible to have a good answer to a bad question? Such as one that explains where the questioner is confused and helps to enlighten them? – Time4Tea Jul 28 '18 at 23:46
• @Time4Tea broadly speaking, i think bad questions with answers are damaging to the reputation of the site. It also seems to me that your understanding of bad is very likely different than mine, so before we go further, state what you mean by bad and I'll let you know how far from me you are – Kyle Kanos Jul 29 '18 at 0:16
• I'd suggest these categories for bad answers: (1) Bad Physics or contains bad Physics (2) Doesn't answer question (3) Hard to follow (not just because the topic itself is hard) (4) Shouldn't have been offered (perhaps because it's a full answer to a homework-type question). What's more I think that down-voters should be required to give a reason, perhaps simply one of these categories. – Philip Wood Jul 29 '18 at 18:35
• @KyleKanos A good answer to a bad question is one that gets the questioner to rethink their question. This is important in physics, is it not? To learn how to make our curiosities well-defined? And to help others do so? – N. Steinle Jul 30 '18 at 3:47
• It may be relevant to note that there's a badge called 'Reversal' which involves posting an answer with a score of more than 20 to a question with a score less than -5. – user191954 Jul 30 '18 at 6:57
• @N.Steinle I wrote ...state what you mean by bad [question] and I'll let you know how far from me you are I don't care how you define a good answer to bad question, I asked OP (Time4Tea) to define what their understanding of a bad question is. – Kyle Kanos Jul 30 '18 at 9:58
• @KyleKanos the sort of 'bad question' I had in mind, where it is possible to have a good answer, is one where the questioner seems to have a fundamental misunderstanding of basic physics. So it depends if you would include those questions in your definition of 'bad questions'. I also think it is possible to give a good answer to a duplicate question - the answerer may not be aware the question is a dupe, and their answer may be better than those given to the original question. I like Chair's comment above - was going to mention that myself. – Time4Tea Jul 30 '18 at 17:19
• @Time4Tea that's not a particularly useful definition of bad question, considering literally every question on this site is by someone who is confused about something. – Kyle Kanos Jul 30 '18 at 18:41
• The title suddenly struck me as kind of extreme. Can you change it to 'Are people voting less on Physics SE?' or something to that effect? Ideally I'd do it myself, but I'm not sure if the exaggeration is deliberate. – user191954 Aug 1 '18 at 15:26
• Not sure if that could be related, but we've seen a decline of voting on SO a lot after the recent "Welcoming Wagon" thing. There has been a decline in downvotes, leaving plenty of... very crappy questions all over the place. The real healthy vote is the down vote, on bad posts, not the trigger happy "you dared to ask an unresearched question, how brave of you, have an upvote" one. – Félix Gagnon-Grenier Aug 1 '18 at 17:45
• Guilty as charged. I've noticed a marked reduction in the quality of the questions as of late. Too many fly-by-night questions by newbies who ask one question and never revisit the site, too many questions by people asking us to do their homework, too many pop-sci physics woo questions, too many duplicates, and far too few quality questions. I at first downvoted heavily when I started noticing this trend. Now I rarely even open questions with goofy titles asked by those with a very low reputation. I can't vote if I don't even look. – David Hammen Aug 2 '18 at 23:26

These statistics are included in the site analytics page (25k+ rep). There is generally a lull in posting rates in the summer and over the winter holidays, but the trend on the votes is less marked. Here are the (weekly-averaged) post and vote rates over the past three years:

The data before 2015 do show a year-on-year increase, which has since stabilized. I normally don't set much stock on data taken roughly between June and August, as the summer does tend to see much lower traffic (and indeed the traffic tends to show stronger seasonal variations, but a recent data outage hasn't been fully recovered), but the mid-semester high point in voting last March does seem to be below the high-water marks from 2017 and 2018.

I'm not sure whether this is enough evidence to be concerned about, but it isn't conclusive evidence that everything is all right, either.

(If people want to play around with the data, it is easily available from the Data Explorer - here is a query to get you started, though it's probably easiest to analyze offline.)

• This show the number of votes as relatively flat, but we know that the cumulative number of users and posts are increasing. As users (and posts) increase, votes should increase too, not stay flat. Seems fishy. – Brock Adams Jul 28 '18 at 9:50
• An important thing to note is that although the number of votes per week has remained approximately stable over the last few years (according to those charts), the number of posts which exist, and hence the number of posts receiving those votes has increased, so the votes are more thinly spread. Of course, to a great extent, the older posts won't be receiving many votes now; most votes would be cast on new posts... but it's still something worth considering. – user191954 Jul 28 '18 at 10:27
• @Chair That information is accessible to analysis via SEDE. My gut feeling is that it won't play much of a role, but you never know. – Emilio Pisanty Jul 28 '18 at 18:17
• It seems to me from the lower chart that the voting rate is perhaps 10-15% lower than the same month (June) two years ago. The overall trend doesn't look too rosy. – Time4Tea Jul 28 '18 at 23:55
• @EmilioPisanty Would a query to find number of upvotes on posts that're less than 1 week old at the time of the vote solve that? – user191954 Jul 29 '18 at 16:25
• It looks like noise to me. – Volker Siegel Jul 29 '18 at 19:33
• @VolkerSiegel Luckily there are tools to extract trends from noisy data. For the voting data, the peak monthly averages for 2018 are definitely some >5% below the 2017 and 2016 maxima, probably at the level of $1\sigma$ or so (i.e. not something that rates at CODATA, but nothing to sneeze at in any data that deals with human behaviour). Throwing up one's arms at the slightest hint of noise is one way to deal with such data, but it isn't necessarily a helpful one. – Emilio Pisanty Jul 29 '18 at 19:36
• There is clearly strong noise in the short term - which is per week. The noise per frequency is still high in low frequency. Look at a chart over less time - you will see wild variations. Looking at a large range you see a - You will need to go back to see all. And from very far you see a straight line. It only depends on the difference of the resolution of the chart and your eye. – Volker Siegel Jul 29 '18 at 19:54
• @VolkerSiegel Sure. If you blur your eyes enough then there is never any signal, it's all just noise on top of a flat straight line. I don't see how it's useful, but it's one way to see things. – Emilio Pisanty Jul 30 '18 at 9:08
• @EmilioPisanty it is important because the interpretation is directly controlled by then author! He can reduce trends, or enhance them. The problem could be solved by publishing the frequency used. But that is too hard to understand. – Volker Siegel Jul 30 '18 at 11:00
• @VolkerSiegel The data is transparently reported (weekly averaging) and the basic data set is open to the public for inspection on SEDE if you want to do your own analysis. (But that does not seem to be what you want - thus far you've only advocated against any sort of analysis.) If you want to continue trolling, please do so elsewhere. – Emilio Pisanty Jul 31 '18 at 8:45
• I used your SEDE query to gather data from 1-1-2011 to 22-07-2018, and then plotted 120-day averages on excel. Looks like there's been a noticeable decrease in average votes per day since 2016. Excel file on google sheets. Image on imgur (random x axis, and the last dot shows the average over 25 March '18-22 July '18, inclusive) Also, the peaks are usually in the april/may-august phase, which surrounds the US summer break. Strange. – user191954 Aug 2 '18 at 5:37
• @Chair That graph sure looks like it could use some error bars. – Emilio Pisanty Aug 2 '18 at 8:11
• Hehe more importantly it could use a label for the x axis... but I get the point that it doesn't indicate the deviations or anything, which makes my claim look a lot more statistically certain than it actually is. But the fact that you can still clearly see some annual trend shows that it isn't concealing too much, but it isn't very noisy either. Anyways, here: imgur.com/a/2GMbNP7 has standard deviations. I think it's ok that the errors are huge... if you think of it, even one person casting 40 votes in a day is going to be a pretty large signal in the raw data. – user191954 Aug 2 '18 at 9:26