The title refers to one of the most common misues of statistics by people with a preset political agenda. An excellent example of this phenomenon appeared in the Swedish media yesterday evening when the lead headline on the Swedish national broadcaster’s text-TV service was the highly provocative “Fler kvinnliga lärare ger lägre löner” (More female teachers yields lower salaries).
http://svt.se/svt/jsp/Crosslink.jsp?d=22620&a=1472087&lid=senasteNytt_275214&lpos=rubrik_1472087
The article goes on to describe a study carried out by the Swedish Teacher’s Association (Lärarförbundet) which compares salary growth in the teaching profession over the last 50 years with that of engineers and business graduates. The chart shows clearly that teachers have performed poorly in this company with their average salary having gone from being 25% higher than in the other groups to being 25% lower today. It is noted that during the same period the percentage of female teachers has increased dramatically. The authors of the study thus conclude that teachers’ salaries are lower because the profession is female-dominated and that this is clear evidence of so-called “structural discrimination” in society.
Let’s assume for the sake of argument that the figures themselves in this study are accurate. The important point is that, even if this is so, the conclusions drawn are completely unwarranted. All that has been observed, to use the statistical parlance, is that two quantities
A = teachers’ salaries as a % of that of engineers/economists
B = % of female teachers
are negatively correlated, that is, A has decreased while B has increased. In general terms, there are three possible conclusions one can draw from such an observation (reality is usually somewhere in between) :
1. The decrease in A is caused by the increase in B
2. The increase in B is caused by the decrease in A
3. The two events are in fact entirely unrelated and are caused by other factors not present in the study. There might even be a common factor which causes both.
The study’s authors have opted for conclusion #1 : the headline on Lärarförbundets own website
http://www.lararforbundet.se/web/ws.nsf/documents/00364DCA?OpenDocument&menuid=00326A1C
makes it clear that SVT haven’t misinterpreted things. This is, at the very least, plain awful thinking. Conclusions 2 and 3 are just as consistent with the data and cannot be ruled out without further investigation.
Here’s an alternative hypothesis as to why teachers haven’t kept pace with other professions in Sweden in recent times. It could be because Swedish teachers are not as competent professionals as they used to be. A great dumbing-down of the educational system took place in Sweden during the 1970s which in particular involved a drastic reduction in the academic qualifications required of teachers. This process has slowed down in recent years but, unlike in some parts of the US, still has not gone into reverse. It’s a long story, and well known to everyone involved in education in Sweden. But as one concrete piece of evidence for this hypothesis, I give you this very study I’ve been talking about. If even the head of Lärarförbundet, Ms Eva-Lis Siren (who is quoted by SVT) is so academically incompetent as to not understand the difference between correlation and causality, then it doesn’t say much for her colleagues as a whole.
I’ll finish with a final piece of speculation. Lärarförbundet are openly using this study as an argument that teachers deserve to be paid better. So could it be that the whole point of the exercise is to open a negotiating front at a time when the economy is in crisis and cutbacks in public expenditures are being considered by all industrialised nations ?
March 11, 2009 at 7:30 am |
Okej nu jobbar visserligen jag för Lärarförbundet och får väl anses partisk och det ska sägas att jag håller med dig i stort. I det politiska spelet övertolkades den här undersökningens reslutat. Samtidigt innebär inte det att den politiska poängen (att lärare är för dåligt betalda) är fel. Och det finns det ett par punkter i ditt resonemang som är ohållbara.
För det första är det bara i teorin som “conclusion 2″ är rimlig. I praktiken är det fullständigt otänkbart att fler kvinnor skulle söka sig till läraryrket som en konsekvens av den försämrade löneutvecklingen. (Kanske de kvinnorna som har en avancerad form av masochism – men inga andra.)
För det andra är du ute på hal is när du använder dig av simpel populism när du ska förklara den “verkliga orsaken” till lärarkårens dåliga löneutveckling. Här har du så mycket fel att vi behöver bryta ner dina argument:
2A – “It could be because Swedish teachers are not as competent professionals as they used to be.”
Om vi bortser från att det här är ett kontroversiellt påstående som inte underbyggs med bevis så är det ändå ohållbart eftersom du inte förklarar vad du menar med “kompetens”. Tar man en traditionell definition av kompetens så har du förmodligen rätt. Men då vill jag påstå att lärare gärna får vara sämre bildade i latin om de istället har fått exempelvis utbildning i vilka möjligheter IT kan ha för elevers lärande om det används på rätt sätt.
2B – “A great dumbing-down of the educational system took place in Sweden during the 1970s which in particular involved a drastic reduction in the academic qualifications required of teachers. This process has slowed down in recent years but, unlike in some parts of the US, still has not gone into reverse.” Jag håller med dig om att de pedagogiska strömningarna under sjuttiotalet var åt helvete fel men de har egentligen ingen koppling till lönerna. Det är ju först under 1988 och framåt som skillnaderna i undersökningen växer. De manligt dominerade yrkena skjuter i taket när lärarnas höjning uteblir. 1988 är ett intressant årtal då det samma år kom en helt ny lärarutbildning eftersom dåtidens lärare ansågs för “flummiga”. Alltså är det inte särskilt troligt att de stora skillnaderna är en produkt av sjuttiotalets tramsande. Jag misstänker att den uteblivna lönehöjningen istället beror på att skolan fick ta stryk under den ekonomiska krisen på nittiotalet. Mindre pengar till skolan gjorde att man inte hade råd att ge lärarna samma löneökning som jämförbara grupper fick.
2C – “It’s a long story, and well known to everyone involved in education in Sweden.” Det här ett festligt påstående eftersom Sverige har ett förflutet som ett av OECDs toppnationer när det kommer till utbildning. De senaste åren har vi visserligen sett övriga medlemsländer komma ifatt oss på flera områden men något tapp är det inte tal om. I matematikdelen av PISA har svenska elever de senaste 12 åren konstant höjt både vårt medel- och medianvärdet. Så i princip kan man säga att du överdriver dina slutsatser och förklarar dem med att påstå att alla håller med dig…
För det tredje avslutar du med att kasta hästbajs på lärarfacken för att vi fortsätter förhandla löner när det går dåligt för samhället. (Faktum är att jag just nu sitter på tåget till Stockholm för att diskutera möjliga ingångar i nästa avtalsrörelse…) Du skriver: “So could it be that the whole point of the exercise is to open a negotiating front at a time when the economy is in crisis and cutbacks in public expenditures are being considered by all industrialised nations?”
Det bristande i ditt argument är att långt ifrån alla länder drar ner på skolans budget i kristider. Faktum är att den dominerande teorin till varför många länder har kommit i fatt oss i internationella undersökningar gjorda under 00-talet ligger i just den ekonomiska krisen som vi hade på nittiotalet. När andra länder satsade på skolan så blev skolan i Sverige offrat på altaret. Då går det som det går. Man får de resultaten man betalar för liksom… Något av det absolut dummaste ett i-land kan göra i en ekonomisk svacka är att försämra kvaliteten på skolan. Det får enorma konsekvenser. Många analytiker menar att vi aldrig hann återhämta oss fullt ut efter nittiotalets kris innan den här krisen slog till. Därför behöver vi både satsa mer pengar direkt i skolan men vi är också i akut behov av lärarlöner som gör läraryrket attraktivt.
//Jacob
March 11, 2009 at 5:32 pm |
Various remarks :
1. One thing I didn’t think of with my original post is that even the conclusion that there is a strong correlation between salaries and % of women is not even justified unless the study also presents a detailed graph showing the evolution of the latter. Otherwise all you have is two data points, one 50 years ago and another now. I can’t find the study any more on LF’s homepage, so I don’t know if such a stronger correlation exists.
2. The way you interpret “conclusion 2″ indeed makes it sound far-fetched. But a more subtle interpretation is far less so. Namely, one can hypothesise that women have a statistically greater tendency to prioritise non-career activities and thus are more likely to end up in lower paying jobs, the competition for which is generally lower and which in general demand less single-minded ambition.
3. Based on the data, the most reasonable conclusion however is #3, I think. Here’s an analogy : 50 years ago there were zero female professional footballers in Sweden. Now there are certainly some. During the same period, the average inflation-adjusted salary of a professional footballer has skyrocketed. Is it reasonable to conclude that this is because there are more women in the profession, and that this is evidence of structural discrimination against men ? Clearly not, for 2 basic reasons : (i) there are obvious reasons for the salary increases which are completely independent of the gender profile (ii) within the profession one can still observe a clear gap between the salaries of men and women, which also can be simply explained by other factors. (i) is a type 3-conclusion. Where (ii) leads us is towards the realisation that the only way one can hope to give clear evidence of structural discrimination is if one can exhibit salary differences for people doing exactly the same work. And even then, there may be other factors which can explain the difference but which are hard to quantify, certainly harder than in the case of footballers.
There is some evidence of structural discrimination at this level, but it is not very strong. During the 2006 election campaign, after hearing Gudrun Schyman (a.k.a. Sweden’s most ridiculous politician) claim on TV that women earned only 73% of what men did for the same work (she explicitly said the words “för lika arbete”), I read a detailed report which SCB had compiled on this issue. GS was of course talking nonsense : the 73% figure just refers to average salaries over the whole society, period. Of this 27 point differential, there remained 7 percent which SCB could not account for by differing professional profiles. So this leaves a ceiling of about a 7% pay differential for “samma arbete”. But this is not yet a proof of any discrimination. Commonly hypothesised factors which SCB could not adjust for because they are hard to measure include (i) maybe women are worse salary negotiators (ii) maybe employers have higher expectation that a woman employee will later give up her job to raise a family etc, and thus don’t offer them quite as good conditions.
4. The entire point 2B in your post seems to me to have zero relevance to the statistics presented in the study. Look at the 3 graphs on the SVT website again : from 1958-68, the löneutveckling in all three categories is the same. Then thngs slow down for the other two groups, but for teachers there is a catastrohpic FALL in real salaries. This fall continues at a slowing rate until 1988, when teachers’ salaries pick up again, at the same time as there is a big improvement for the other groups. From 1998 onwards, the 3 groups are all improving at approximately the same rate again.
Clearly, while something significant happened in 1988 for the other groups, the really significant events for teachers happened in the years after 1968. There is thus ZERO evidence for your theories. There is some evidence for my hypothesis that things went up the spout in the 1970s, though it seems that something was already going seriously wrong in the late 1960s.
5. You are quite right to point out that I was not being at all rigorous in my first post. But I did not intend to be, and I stated clearly that I was only presenting hypotheses. As I stated in point 5., I think this study in fact gives my hypothesis far more plausibility than any of yours. But, of course, I haven’t proven anything. Indeed, because of the correlation/casuality issue, one cannot prove anything unless one has a very detailed data set on which one has done a very careful series of hypothesis tests. If anyone has done so on the issue of teacher’s salaries, I would love to know. This particular study would seem to be anything but serious, at least the way it’s been interpreted.
6. I don’t want to comment on the larger issues you raise in your last paragraph now. I am currently reading a book on developmental economics and will be making a blog post on that topic soon.
March 11, 2009 at 6:50 pm |
One footnote to my last remarks :
I can think of an entirely different hypothesis to explain the various löneutvecklingar which is based simply on market forces of supply and demand.
Apparently, universally funded higher education was introduced in Sweden in the mid 1960s. One can then suppose the following chain of events :
1. Numbers of university graduates increased dramatically in the late 1960s, especially numbers of women
2. Thus there was a glut of supply in more highly skilled professions from the late 1960s onwards.
3. This caused downward pressure on salaries across the full range of professions, but especially in those which proved more attractive to women, since there the supply glut was largest.
I do not have the data to back this up – it may be complete nonsense. But this is the type of explanation which I prefer as a default, since it doesn’t require any conspiracy theorising. It also suggests how the lower salaries of teachers could in fact be caused by the higher % of women, without this implying discrimination.
March 25, 2009 at 10:15 pm |
Hi!
It would be interesting to know if the number of females (in percent) among engineers/economists has increased too. If this is the case, why hasn’t this yielded in lower salaries?