I have been buying masala dosa from a local shop since the 2000s, and at some point I started noticing the price. Not tracking it properly, just noticing it, the way you notice a friend’s hairline.
Here is what I can piece together.
| Year | Price of one masala dosa |
|---|---|
| 2005 | Rs 20 |
| 2010 | Rs 30 |
| 2015 | Rs 45 |
| 2020 | Rs 60 |
| 2026 | Rs 75 |
Five data points, no receipts, entirely from memory. Naturally the correct response is to treat this as a serious economic instrument. I present the Masala Dosa Inflation Index.
For a compound annual growth rate we want
$$CAGR = \left(\frac{P_{end}}{P_{begin}}\right)^{\frac{1}{n}} - 1$$
Rs 20 to Rs 75 across 21 years gives 6.50 percent per year.
The dosa is up 275 percent. It costs 3.75 times what it did in 2005. Doubling time at this rate is about 11 years, so the dosa quietly doubles roughly every eleven years, which is a fact I would rather not have learned.
India’s official retail inflation has averaged roughly 6 percent a year over this stretch.
The dosa says 6.50 percent.
Five half remembered prices from one shop, and the answer lands within half a percentage point of the national figure. No CPI basket, no consumption weights, no methodology, no effort. Somebody set a price based on the cost of onions and rent and the general feeling of the year, and it tracked the economy.
If Rs 20 had grown at the official CPI rate since 2005, you would need Rs 68 today. The dosa asks 75. So it has outrun the official number by a small, dignified margin.
Split the index into segments and something reassuring appears.
| Period | Rate |
|---|---|
| 2005 to 2010 | 8.45 percent per year |
| 2010 to 2015 | 8.45 percent per year |
| 2015 to 2020 | 5.92 percent per year |
| 2020 to 2026 | 3.79 percent per year |
Dosa inflation is decelerating. The 8.45 percent era is over. The most recent stretch is a very mild 3.79 percent per year, which for a period containing a pandemic is remarkably restrained.
What Rs 100 gets you:
| Year | Dosas per Rs 100 |
|---|---|
| 2005 | 5.00 |
| 2010 | 3.33 |
| 2015 | 2.22 |
| 2020 | 1.67 |
| 2026 | 1.33 |
In 2005 a hundred rupee note could feed a small group. In 2026 it buys one dosa and change that cannot buy a second one. Somewhere in there the hundred rupee note stopped being a plan and became a formality.
Assume 6.50 percent holds forever, which it will not.
| Year | Projected price |
|---|---|
| 2030 | Rs 96 |
| 2040 | Rs 181 |
| 2050 | Rs 340 |
| 2075 | Rs 1,639 |
| 2100 | Rs 7,904 |
The dosa crosses Rs 100 around 2031.
By 2050 it is Rs 340, which means anyone hearing about the Rs 20 dosa will react the way I reacted to stories about the Rs 2 bus ticket, with polite disbelief.
By 2100 a single masala dosa costs Rs 7,904. Not a plate. One dosa. At that point it is less a breakfast item and more an asset class.
This projection is of course meaningless, since it assumes a fixed rate for 74 years. But it is meaningless in a rigorous way, derived from real prices, which is the respectable kind of meaningless.
Two things.
Inflation is invisible up close and enormous from a distance. Every individual increase here was five or ten rupees. Not one of them was worth reacting to. None of them felt like an event. Add up twenty one years of increases nobody noticed and you get a 3.75x. That is compounding doing its usual trick, which is to be completely unremarkable right up until you look at the total.
Duration beats precision. My methodology here is genuinely terrible. Five prices, no records, pure recall. But the time span is long, and a long sloppy record turns out to be enough to recover a country’s inflation rate to within half a percent. Twenty one years of noticing carelessly beat two years of measuring properly. That is not a compliment to me, it is just what happens when $n$ gets large.