9% growth… from selling the country

“Rajeev Chandrashekhar, a Rajya Sabha MP and former FICCI president, has pointed out that agriculture has grown at a dismal one per cent and manufacturing at no more than three per cent. The so-called miracle has been achieved through phenomenal growth in mining, real estate, construction.” So says Neelabh Mishra in his article in Outlook, titled The Banana Sheikhs.

Read this with Jagdish Bhagwati’s lecture The Unfinished Reform Agenda. The good professor says nothing about the sources of growth, even while  he rages against Indian novelists being allowed to write on the economy.

When ordinary people cannot see the growth they are supposed to be enjoying, who do they turn to, economists with figures or novelists with facts?

Broken families and rich individuals

In the video RSA Animate - Crises of Capitalism, David Harvey, a Marxist says that the present crisis has everything to do with fall in real income per family in the Western world; in this one, Crisis of Capitalism, The Critique,  someone debunks Harvey by pointing out that income per capita has increased and that the fall in income per family is simply because there are more families now, that is, if a population of 100 were split into 25 families of average size 4 (persons per family) 30 years ago, now that same population is divided into, say, 50 families of average size 2.

I have heard the same explanation from the Kublai Khan of capitalism, Jack Welch.

It looks too easy to be right.

First, where is the data? Let’s say fewer people are getting married in the West these days. Does that also mean that the size or nature of the family unit, on average, has changed drastically?

Second, if the income per family has dropped, why should one not worry about it? Doesn’t the amount a person spends, and saves, depend enormously on whether he or she is in a family?

Just take rent or mortgage. Suppose a family of four spends x by living under one roof (average spend per person = 0.25x); and a pair of divorced parents with the two children living with their mother spend 1.2x (father’s rent = 0.4x; mother and children’s rent = 0.8x; average spend per person = 0.3x, 20% more than the average for a 4-member family). Does that not make a significant difference?

Plus, the mother’s income may be less than her married counterpart’s because she has more on her plate (no-one to share her load with).

The father, on the other hand, may be spending more on conspicuous consumption than his married counterpart does.

In fact, both parents may be spending more on sex (wining, dining, gifting, grooming to entice mates, or straight cash) than they would have had they been in a family, where sex it is essentially a bonus (free gift?) of family life.

And while married parents (or parents who operate as a family in spite of not being married) may save to provide for the future, parents who do not operate as a family may, for financial and psychological reasons, save little.  

I mean, the word income has very different meaning when applied to a person than when it is applied to a company (where it means profit). So why don’t Western commentators take that into account?

TV succeeds where all else fails. Or does it?

In the introduction to Super-Freakonomics there is a section on how TV is rescuing Indian women.

Before they come to the boon of TV they give some statics and quote a couple of anecdotes. Then they inform that most government schemes for women’s uplift to have proved ‘complicated, costly, and, at best, nominally successful’.

So what was successful? TV.

Two American economists, Oster and Jensen, found that out ‘by measuring the changes in different villages based on whether (and when) each village got cable TV’ as the unlikely saviour rolled out over the Indian countryside.

The wording is important. Hence I quote: “The women who recently got cable TV were significantly less willing to tolerate wife-beating, less likely to admit to having a son preference, and more likely to exercise personal autonomy.”

After some speculation on the reasons behind this sea change and the veracity of survey the economists’ initial findings on women’s attitudes was based on, the book continues, “Rural Indian families who got cable TV began to have a lower birth-rate than families without TV. (In a country like India, a lower birth-rate generally means more autonomy for women and fewer health risks.) Families with TV were also more likely to keep their daughters in school, which suggests that girls were seen as more valuable, or at least deserving of equal treatment.”

There are two problems with this story. First, it’s old. We have heard a different version while growing up, in which the transistor radio proves to be the most effective method of family control in otherwise entertainment-starved poor families.

Second, it’s very likely that this research mistakes effect for cause and vice versa. Is it not possible that progressive households and villages got TV sooner than regressive villages? The authors do not tell us if there were attitudinal and behavioural differences before any of those villages had TV, and if being able to afford TV may have been the effect, not the cause, of women’s liberation.

Take a different, but closely related, change. In my generation the Indian educated middle class underwent an enormous population contraction.

I don’t have figures but I do have plenty of anecdotal evidence. Only one piece will suffice for now: My father has seven siblings; my brother and I have seven cousins (from our father’s side).

Now, did TV do that or the fact that my father and his siblings are all college graduates married to college graduates?

For that matter, one wonders if it has ever occurred to any economist to research if TV benefited women in the West as it supposedly did in India?

Just as TV didn’t come to the Indian countryside all at once, it didn’t spread over the West in one shot. The data on launch of TV and women’s development shouldn’t be very hard to find. But would anyone even think of linking them for developed countries?

I admit I haven’t read the original paper, but so wouldn’t most of the readers of the book. In such case, shouldn’t the authors have mentioned that possibility? Doubtlessly it must have occurred to them.

You can’t win

I’ve been reading and listening to Jeff Pfeffer on evidence-based management. Basically, the good doctor wants managers to look at the theories, instead of mearely going by beliefs that sound ruthless enough, i.e., in line with with Econ 101, e.g., ‘fire the losers, give stock options to winners, etc’.

Two problems come to mind immediately. Managers may not know how to look at evidence. Those who come from the Arts, may not know any statistics, and may even be scared of numbers. Those who come from the Sciences, while somewhat better equipped, may over-estimate the power of numbers… mistake data for information. Men and women are not screws and nails. Companies are not bridges and towers. Markets are not computer models.

The second problem is that the evidence at hand may be misleading for complex systems. Let’s take, say, stock options. Apparently, there is no evidence that it makes any difference. In some cases, it is counterproductive. So what? The question is not ‘What does it do in general?’ but ‘How do the people who do it right do it?’

(Let me take an extreme example to amplify: There is no evidence that keeping accounts will make your business prosper. All businesses keep accounts of some sort; most fail. Hence, we should conclude that keeping accounts is immaterial to business success.

But we don’t. Instead, we try to figure out the best way of keeping accounts, because commonsense tells us, without the need of any research, that accounts should matter a great deal.)

I have the same problem with this thing as I had with The Bottom Billion by Paul Collier, namely that applying data analysis may be as much as an over-simplification as not applying any. A diverse, representative sample may be as wrong as a sample of a single failure or a single super-success.

Why movie stars pay the most tax

The film ads proclaim, “Rs 120 crore collection!” The seats are empty.

What’s up? Yesterday’s Times of India offers hope out of the mystery.

According to an article titled India makes almost as many films as US, Japan & China, an average Indian movie ticket costs $ 0.5 while an average American ticket costs $ 7.2 and a British ticket costs $ 9.5.

Now, Hindi, Tamil and Telgu films collect a good amount from overseas. A few reasonable assumptions and you have:

	Here	US	UK
Per ticket	0.5	7.5	9
Seats per hall	400	150	150
Shows per week	28	28	28
Weeks' stay	4	2	2
No of halls	100	30	20
% filled	15.85%	6.16%	4.25%
Seats filled per show	63	9	6
Tickets sold	4,50,27,370	1,43,630	45,455
$ earned	2,25,13,685	10,77,224	4,09,091
Total in $	2,40,00,000
in Rs	1,20,00,00,000

 

Of course, films are imported elsewhere too. The Middle East must be a big market, at least wherever movies are allowed. And there’s Southeast Asia.

And I refuse to believe that the average ticket price commanded by a big movie in its first few weeks is as low as $ 0.5 (Rs 25). Surely they make much more in the multiplexes.

On the flip side, one wonders if Indian films get 4 shows a day throughout the week outside India. But we don’t need the details now.

The point is that with a weak rupee, a hit rate of1 in 25 should be good enough to turn any movie into a blockbuster, and Indian moviemakers need never look for non-Indian (Chinese, Western) audiences because Indians abroad are enough!

No wonder stars pay the highest taxes: They have so much to spare. The tragedy is that almost everyone else connected to movies – technicians, bit players, ushers - seem to be making a pittance. Why?

Minority is ok

Many tears are shed for callous citizens who never vote. Someone gets elected in spite of only 60% or so voting. Worse, he may have the support of the minority of those voters.

I can't understand what the hullaballoo is all about, because a little thought shows that the damage, if any, is only to the losing contenders. For the citizens, both non-participation and election with minority support are immaterial.

For why do we need to know each voter's preference when a representative, random sample should give us a very good idea of what people want? In a populous country like ours, even the voters in a municipal election will yield a sample size that's large enough to give a truthful picture of the population's preferences.

It may be argued that the voters do not usually constitute a representative sample. For instance, poor people are more likely to vote. So the richer classes would be underrepresented.

However, it can just as easily be argued that those who stay away have reasons for doing so. If an electorate (those who can vote) is 1% rich and 99% poor, does it really matter if its electors (those who do vote) are 0.75% rich and 99.25% poor? Is it unreasonable for the rich to believe that their concerns won't count in the elected body? Or that they will have to use something other than votes to make themselves heard?

In fact, it is the insistence on making every vote count that keeps many away from politics. Let me explain. Let's say the poor decide to vote en block for a certain candidate (and agenda). They vastly outnumber the rich. So some lazy poor decide to stay at home believing, rightly, that their absence will make no difference, more so because the rich will be underrepresented.

The rich, on the other hand, know that for their votes to count, they must turn out in full force. They are, to begin with, marginal. Now, let's say Rich Guy A says, "What is the probability that my neighbour will vote? 15%. That's slim. And the chances of my other neighbour's voting? Well, 20%. That's bad. I wanted to vote, and I know it's my duty, but thanks to the callousness of my class, my vote won't count. So, why should I waste my time? Let me watch a movie instead."

His neighbours have more or less the same thoughts and stay away, not just from the polling booth but also from the entire election process. In short, they are doubly damned: first, for being a minority; second, for thinking cynically.

Of course, you cannot have election an electorate formally divided by class, religion, caste, and the rest. That would divide our already fragmented society even more, and we'd forever be at each other's throats. So, we better not insist on representative sampling. But we can insist on random sampling, hoping, again very reasonably, that the sample that yields will be representative as well.

This may, who knows, bring the added benefit of improving overall interest in politics among the previously 'doubly damned', who should get rid of their cynical reasoning (for staying away) once they don't have to bother about 100% turnouts any more.

Let's look at the problem of getting elected by a minority. This is slightly complex. But let's take an imaginary situation where there are 5 candidates; 3 get 20% each; 1 gets 19% and the winner gets through with 21%. Only about 1 in 5 voted for him; 4 in 5 didn't want him. Bad? Yes, if you want a simplistic answer. However, if you take a step back, it hardly looks horrible.

The split of the vote says that while none of the candidates was very good, none were very bad either ('good' and 'bad' merely being measures of ability to align oneself with the electorates' preferences). So, how does it matter to the voters who won!

Perhaps no important legislation was to be decided in parliament. So the choice of legislator became a popularity contest. While who proves most likable matters to the contestants, it makes little difference to the public.

Perhaps the legislation was important, but debating it wasn't. For example, let's say a bill proposes, "No person can be executed without proper trial." It's critical, but scarcely debatable.

Or take a verdict that goes thus: Winner-31%, runner up-30%, spoilers-29%. Wouldn't we like voters to have indicated their second and third preferences in such cases? Such methods are actually followed in quite a few elections. But it may be quite unnecessary.

Because the moot question for the voters is not Who won? Instead, it's What happened because he won?

If you got through by the narrowest of margins, and wanted to improve your (or at least your party's) chances next time around, what would you do (Only a mad electorate will elect someone who doesn't want to be re-elected)? Would you become very biased? Not unless you can indulge in some large scale ethical cleansing. If you can't, you'll look after everyone's interests. In either case - can or cannot commit mass murder – your getting a minority vote is, by itself, no major tragedy. (If the winner can commit mass murder, elections and elected bodies have long ceased to make any difference.)

The complacency about winners with a minority of the votes becomes more justifiable these days because many of our governments are post-election alliances. The very fact that politicians can get into bed after blooding each other's noses in public shows that they were fighting for power (which concerns them) and not principles (which concerns us).

Actually, in a democracy with a diverse population, it's high turnouts and overwhelming majorities that we should be worried about.

When everyone turns out to vote, they are probably worked up about something. Either riot is in the air, or revolution. And an overwhelming majority can only be a goon's. 

Even more admission nonsense

This follows from More Admission Nonsense (http://directindia.blogspot.com/2008/08/more-admission-nonsense.html). Apparently, the normalisation math goes thus: 'The average of the top 10 scores in a particular board is calculated; the scores of a student from the same board is then divided by this average; the number is then multiplied by 100 to arrive at the student's normalised score.'

At least that's what yesterday's (26 Aug 08) Times of India says.

If there is any logic to this, it escapes me.

The top students of any board score almost full marks these days. So the whole exercise effectively becomes multiplying a student's score by 0.99987 or some figure very close to 1. 

Anyway, since when did these outliers become representative of a board? Shouldn't they be looking at the median, mean or mode, or some such central measure? Why don't they simply ask boards to give all their marks, whereupon one call easily calculate each student's percentile (with a OTS software)? What do other boards do?  

Business schools' annual reports

Business schools professors shout themselves hoarse about the importance of standardised scores (the only score that matters in marketing is Net Recommenders) and scoring methods (to compare Company A's profits with Company B's, one must first be sure that both were calculated in, more or less, the same way). 

Very right.

But shouldn't they begin at home, by agreeing on standard methods of measuring success, and even go a step further to issue fairly detailed guidelines of how these figures should be used? For instance, students interested in a career in marketing should look at scores X, Y and Z, whose weights are a%, b% and c%, respectively. 

Considering that most school sites are very similar, education agents have self-interest in recommending one course over the other (How do we know they don't get commissions and only fees?), and magazine ratings are unreliable (No 1 in list A doesn't exist in list B), and all sorts of rumours make their rounds, candidates cannot be blamed if they are totally confused about where to go. 

On the face of it, it is not in the school's self-interest to help, obviously, given the immense asymmetry in information. But do we have conclusive proof that this is indeed so?  

It goes without saying that true excellence cannot be compared, even less quantified. However, that is as true for the marketplace as the academic ivory towers. Why shouldn't the ivory tower dwellers practise what they preach?

On second thoughts, two scores are readily available: first, the application fee, and second, the tuition fee. The number of publications by students and faculty may be useful too, particularly for the more academically inclined. But are these readily available and easy to collate? Can't college associations make it mandatary to publish certain data, and make the data accessible to everyone with, say, a GMAT score? 

By way of the movies

Freakonomics has an interesting piece on research into baby names in California. In a nutshell, the research finds poor, under-educated, black single mothers 'act black' while naming their new children, which may be indicative of the way they'll bring up their babies.

It also says names travel down classes, that is, a name becomes popular among the rich before it becomes popular among the poor, by which time the rich decide that it's become 'down-market' and abandon it, unless they want to project a 'regular guy' image.

It'd be interesting to find out the role of popular culture, especially movies and tv shows, in the names' transition. Do names that become popular among the rich also become popular among screen characters, via whom they reach the poor? Or does the janitor pick up names from the bosses' doors?  

(Do screenwriters name characters after their friends or their friends' children? If it is the former, one should expect a reallife-to-screen shift to take a few decades; if it is the latter, the gap should be far shorter.)

A big question is why can't we get such interesting research and books in India? Is it because we are too busy making a living? But interesting research provides a good living? Or are we plain stupid? Or made stupid by our rote-and-rot education system?

Size of direct marketing industry in India

I did some Femi calculations, which are at http://spreadsheets.google.com/pub?key=pM63eRG2NCQi6YsR03FUpIA. The estimate is $ 151 billion! I'm going horribly wrong somewhere.

Ugly music

A big filmstar is going to host another one of those horrible reality shows where, probably, families will challenge each other in who sings most off key. The ads are up across the city. I noticed two.

Each features a Hindu family and a Muslim family.

By chance? Unlikely.

India is 80.5% Hindu, 13.4% Muslim. If you pick 4 Indian families at random, the chance that you'll end up with two of each religion is 1.16% (sq0.805 X sq0.134).

Having done this 50:50 picking, your chances of following up with a 50:50 pairing is 2/3 (H1, H2, M1 and M2 can be paired thus: H1M1 & H2M2, H1M2 & H2M1, and H1H2 & M1M2).

Combining, we find that there's only 0.78% (0.0116X 2/3) probability that filmstar & Co ended up with the Hindu vs. Muslim pairs by chance.

I have assumed there are only two ads. If there are more, and they have inter-religious pairing too, then the chances would go down further.

I cannot but see these Hindu-Muslim ads with jaundiced eyes. Ostensibly, they invite the country to make music together. Underneath, one suspects, lies a sinister if silly intent.

33% better than random

Every book on the application of statistics in database gives examples of how their model overtook a random sample by so many percentages.

I am yet to come across the copywriter who says, “My mail pack did 35.78% better than the one fashioned by a one-eyed monkey banging away on his keyboard.”

What makes beating a random sample so commendable is beyond me.

The killer average

Let's say you are running a loyalty programme where one of your main objectives is to use points to bring down discounts. Five months into the programme, you discover that average discounts outside the programme are marginally more than those in the programme (let's say, it's 27% inside the programme verses 29% outside the programme).

Worse, when you add the deferred discounts you have to give away as points (let's say, 2%), the discounts run neck to neck.

Worst, when you put in the expenses on running the programme, it seems quite clear that the programme is, effectively, extracting a larger discount.

Or is it?

Perhaps the transactions inside the programme are very different from those outside it. For instance, sales inside the programme can be of high-price, high-margin items; while those outside be of low-price, low-margin products.

The discounts percentages can be coincidentally equal.

The correct comparison would be of the fraction of the margin given away as discount. Better still, one should ask the simple what-if question: What would have the total discounts been had the programme not existed, assuming everything else remained the same?

This sounds obvious and simplistic, yet every day I see numbers being used with the minimum thought about where they came from.

Which brings me to a more fundamental problem. While finding figures for a nation like ours, where inequalities are far higher than people can ever imagine in developed countries, are agencies careful enough to cast their nets wider for data?

Or would we be better off if, while investigating anything to do with economics and money, we divided the world by borders that combined purchasing power with politics?

No beta, yes risk

In direct marketing texts very little is discussed about sample sizes, and even less about Type I and Type II errors. Admittedly, the latter is somewhat complicated, and business statistics books often recommend that readers go over sections covering it carefully and repeatedly.

‘Complicated’ is, unfortunately, not synonymous with ‘of theoretical value only’. In fact, the opposite is true in this case, because the Type II error is of fundamental importance, as becomes apparent if we step back and ask why we bother about test and sample size in the first place.

We do so, basically, for two reasons. First, we don’t want to throw the baby out with the bathwater. We don’t want to take a test result that is somewhat less than our expectation on its face value. We’d rather use the test to estimate the list characteristic leading to the result, and if that turns out to be acceptable, we’d like to scale up. This is why we worry about a (the probability of rejecting a true null hypothesis).

At the same time, we don’t want to lose money by scaling up when we shouldn’t have. This means we should worry about the minimum response that would be ok. For this we must worry about b (the probability of accepting a false null hypothesis).

Yet, the commonly used formula for sample size completely ignores beta (not only in direct marketing books and online calculators but also in most of the business statistics texts I’ve come across)!

The formula goes like this:

N =	za/22p(1-p)
	E2

Where

z is the value used for the specified confidence level

p is the estimated response (population proportion) and

E is the ± sampling error allowed.

Let’s say estimated response is 2%; the allowed error is ±0.25%; and confidence level is 90%.

Putting these into the template yields a sample size of 8,485.

Let’s see what this means in terms of b if the true response (which we’ll get to know only if we scale up to the entire list) is, say, 1.65%. b turns out to be 23%, that is, 1 in 4!

Humm, that’s bad. There’s a 1 in 4 chance that while accepting a figure between 1.75% and 2.25% as ‘as good as’ a 2% response rate, we’ll actually accept a list with only 1.65% response.

No wonder one of the well accepted rules of the thumb in direct marketing goes: “As a rule, the response rate from a rollout to the balance of the list after a successful test mailing will usually be lower than the response from the test.”

This may be because of a variety of differences between test and rollout conditions. But one thing needs to be kept in mind: If one ignores Type II errors, the success of the test can be very suspect indeed.

A way out could be to use an alternative formula:

N = (	|z0|(p0(1-p0))1/2 + |z1|(p1(1-p1))1/2	)2
	p0 – p1

Where
p₀ is the estimated response

p₁ is the value for which Type II error will be monitored

z₀ is z_a or z_a/2 depending on whether the test is one- or two-tailed and

z1 is z_b where b is the limit on type II error probability when p = p₁.

Let’s see what happens if estimated response is 2%, the allowed error is ±0.25%, and confidence level is 90% (as before); while p₁ is 1.65% and b is 10% (i.e., there is only a 1 in 10 chance – not 1 in 4 as earlier – that we’ll accept a list with a real response rate of 1.65% by performing the test).

We get a sample size of 12,643.

Sure, it’s a 50% increase in sample size. But it may be well worth it if the roll out numbers and costs are far higher than the test’s.

In any case, won’t decisions be better if they were taken with a clearer idea of the risks?

PS: Please excuse the ungainly appearance of the formula. It's the best I could manage using a Word file. Both formulas can be found in the useful templates at http://highered.mcgraw-hill.com/sites/0070620164/student_view0/excel_templates.html

In leaps and bounds…

“In India, we’re skipping technologies.”

“Forget everything. Everything’s changed.”

“Traditional advertising is dead.”

“Direct mail never took off and now it’s too late.”

It drones on and on.

Blanket statements with the surety of the sun’s rising in the east.

And what are they backed with? The ‘fact’ that we now have 147 million mobile subscribers in India. (http://www.coai.in/archives_statistics_2007_q3.htm)

This one statistic is supposed to be symbolic of all that is good and great in emerging India.

Frankly, that ‘fact’ is dubious. I bet we neither understand nor report churn correctly. I guess that 147 figure overestimates the subscriber base by at least 1/3.

Even if it was right, and we threw in all the telephones and Internet connections and what have you, you get a tele-density of 10 to 12%.

Big deal.

Actually, big shame.

It’s like celebrating progress in literacy when every second Indian can’t read, or, for that matter, in economic growth, when one out of four lives below the poverty line.

But I digress.

My main point here is the absurdity of taking one statistic, no matter how impressive, as the basis of a universal judgement. One Indian company buys a western one, and India conquers the word. One Rolls Royse sells somewhere in the back of beyond and we’re all maharajas. One NRI author in English wins a prize and we acquire a voice, at last. The literature of our 25 languages doesn’t count. Never did.

You hear this logical pole-vaulting in meeting after meeting. If a Western person is present, the absurdity attains insanity.

Nobody asks, “What are the facts of this case? Can we see them first?”

70% are businessmen

I won’t have believed this had I not seen this.

One of India’s largest companies did a customer ‘survey’ for one of their new brands. The largest proportion in the finding was 70%: 70% of the buyers who agreed to be surveyed were businessmen.

The upper confidence level for that fraction (at 95% confidence level) is 90%; the lower confidence level is 50%; the figure is meaningless.

Yet this company will use this to take decisions involving tens of thousands.

What can you say about this man?

“He has bought Brand A. What does that tell you about him?” How many times have you heard this question? How many times have you had to made up a wonderful pen picture of this customer, based on this one single purchase and agreeing with brand manager’s reading of who constitute his market?

And how many times have you told yourself that this Sherlock Holmes act is totally absurd?

Holmes would at least have a well-used object, with plenty of tell-tale marks on it, to base his deductions on. All you have is a single purchase. And some completely unsubstantiated assumptions.

Yet brands’ creative and media plans are based on these mental gymnastics.

Is it so difficult to say, “The only thing we can tell about the customer is that he can, most probably, afford this brand. If we have additional data on customers, we can probably hazard a few more guesses. For instance, if we know that 67% of customers are Sindhi grandfathers with four-and-a-half gold-filled teeth, we can say, ‘There’s a 67% chance that our new friend is a Sindhi grandfather with four-and-a-half gold teeth.’ Beyond that we can’t say anything.”?

Why must we know our customers profiles? Why can’t we just restrict our interest to the whys and wherefores of their liking our brand?

We use many brands ourselves: In how many cases do we fit those brands’ (apparent) target markets? Or are we ‘beyond marketing’, non-slot-able, unique, different?

I ‘eurekad’ when I read Caples the first time, but when I started reading again! –

John Caples’s Tested Advertising Methods is a book I will unhesitatingly recommend to any copywriter. As I will recommend his How to Make Your Advertising Make Money and Making Ads Pay.

So what I am about to write is, for me, sacrilege. Yet it must be committed.

The tests that Caples talks about leave a lot to be desired. I was just going through Tested Adverting Methods (5^th edition, revised by Fred E. Hahn [a revision that does it only harm]) again, and found the inadequacy of the data particularly puzzling.

Take the famous example where Caples says changing an ad headline from “Repair Cars – quickly, easily, right” to “Fix Cars – quickly, easily, right” increased response by 20%. He gives no information about what the actual numbers or percentages were, or where the ad came out.

So let’s suspend belief for a while and pick some numbers out of the Web.

The average weekday circulation of a newspaper in the US (whatever that is supposed to mean!) in 1940 was 21,902 and the Sunday circulation was 61,659 (Please see this section of The State of the News Media 2004 report at journalism.org).

Tested Advertising Methods came out in 1932, so here are our assumptions: (a) The circulation was not too different in 1932 (we have no reason to do that, but the data at the site is only till 1940) and (b) Readership was equal to circulation (again, a rather silly assumption, but the purpose of this to explore a possibility and not to prove a point).

The situation we can imagine goes like this:

The ‘repair’ ad could have pulled up to 201 responses

And the ‘fix’ ad 20% more, that is, 241 responses

Without the response rates being significantly different (at a 5% level of significance).

Had the ad come out on an average Sunday newspaper, the responses could have gone up to 209 and 250, respectively, without the response rates being significantly different.

The same complaint can be made against the comparison between “Save one gallon in every ten” and “Car owners! Save one gallon of gas in every ten” where, on testing in a daily newspaper, the latter pulled 20% better than the former.

Another famous example is the one where “Hay Fever” pulls 297 sample requests while “Dry Up Hay Fever” pulls 380, a ‘27% increase’. The increase in response rate (assuming the ads came out in average newspapers on weekdays) could have been between 0.15% and 0.61%.

In quite a few cases, neither response rates nor responses have been quoted; we’re simply told A did better than B.

Now, if the differences in response rates were not always significant, from either a statistical or business perspective, the businesses involved in those testing decisions would not have gained or lost much.

The trouble lies elsewhere, with direct marketing copywriters who believed the ‘tested’ fact that ‘straight and simple always out-pulls the creative’ and put their own careers into jeopardy, because that belief is almost always seen as an excuse for lack of talent.

To all such writers, and to writers who have not yet formed their beliefs, I would recommend this site: Statistics Every Writer Should Know. A little knowledge may be a dangerous thing, but none at all can be disastrous.

PS: I used an Excel template from Aczel & Sounderpandian for my calculations. My calculations are at http://docs.google.com/Doc?id=dd3bjnd7_28rt37t and the templates are available here.

Why Indians don’t read

Why can’t you do any direct mail or, for that matter, long copy ads in India? Simple. Indians don’t read.

But opening your eyes shows you that there are more papers, magazines and books than ever before, and more bookshops. And there is the www. While much of what is written in the world, for work or pleasure, is never read, surely enough is read to sustain the writing, financially and physiologically (Almost all sperms don’t make it, but enough do to make 9 billion of us).

So what’s going on?

While I haven’t got any surveys to refer to except this one, I suppose a simple and possible answer may be obtained if we look at a family’s reading. The calculations are here: http://spreadsheets.google.com/ccc?key=pjtNNMP33DgsJnLoV4Sy1Pw&hl=en_GB. (The logic is not different from the explanation to the GMAT paradox.)

As is apparent, each member’s reading goes up, yet the average, dependent on the number of members, keeps fluctuating – and in two cases, goes down.

It can easily be that, in the larger market, readership is going up, as is each individual’s reading; yet the average reading is going down because neo-literates form larger and larger fractions of the population (while the bibliophiles’ fraction, and their power to influence the average, keeps getting smaller [though their numbers increase]).

In short, simply asking where an average came from could have led to a very different explanation, and decision!

As the chowkidar said

Said the old judge to the youthful civil servant, “When you are a bit older, you will not quote Indian statistics with that assurance. The government are very keen on amassing statistics – they collect them, take the cube root and prepare wonderful diagrams. But what they must never forget is that every one of those figures comes in the first instance from the chowkidar, who just puts down what he damn pleases.”

Substitute ‘buttonholed shopper’ for chowkidar, and that becomes good advice for every survey spouting MBA.