In monetary theory, Goodhart’s Law states that “when a measure becomes a target, it ceases to be a good measure.” That is because people, and even governments and other organizations start gaming the target. Named after Charles Goodhart, Emeritus Professor at the London School of Economics, and a former Chief Adviser and External Member of the Monetary Policy Committee at the Bank of England, who had propounded it, Goodhart’s Law was originally formulated in the context of monetary policy during the Thatcher years. But its utility goes beyond monetary policy in explaining various phenomena where targets are met, but underlying performance is poor.
Goodhart’s Law was cited by Chris Giles in today’s Financial Times (15 May 2020) to explain why UK’s National Health Service focus on deaths in hospitals might have contributed to larger number of deaths in care homes and at own homes, the increase being 19,900 and 10,800 respectively over a five year period. In contrast, Giles says that the target of conducting 100,000 coronavirus tests a day was better conceived. It helped track and isolate those infected by the virus.
In India, there are some early indications of gaming the Covid-19 indicators. For instance, the total number of pandemic cases is a simple and easily understood measure of the extent of the pandemic. But, a low number of positive cases can be achieved by not conducting enough tests. So, a Positive-Tests Ratio (PTR) could be used. If you look at aggregate numbers (as on 14 May 2020, Source: https://www.covid19india.org/), the position for all states with more than 500 reported cases are as below:
| # | State | Number of positives | Number of tests | Positives-Test Ratio |
|---|---|---|---|---|
| 1. | Maharashtra | 27,524 | 2,40,482 | 11.45 |
| 2. | Tamil Nadu | 9,674 | 2,91,432 | 3.32 |
| 3. | Gujarat | 9,592 | 1,24,708 | 7.69 |
| 4. | Delhi | 8,470 | 1,19,736 | 7.07 |
| 5. | Rajasthan | 4,534 | 2,04,243 | 2.22 |
| 6. | Madhya Pradesh | 3,902 | 1,53,139 | 2.55 |
| 7. | Uttar Pradesh | 3,902 | 1,53,139 | 2.55 |
| 8. | West Bengal | 2,377 | 62,837 | 3.78 |
| 9. | Andhra Pradesh | 2,205 | 2,10,452 | 1.05 |
| 10. | Punjab | 1,935 | 47,408 | 4.08 |
| 11. | Telangana | 1,414 | 19,278 | 7.33 |
| 12. | Bihar | 999 | 40,782 | 2.45 |
| 13. | Karnataka | 987 | 1,28,373 | 0.77 |
| 14. | Jammu and Kashmir | 983 | 63,515 | 1.55 |
| 15. | Haryana | 818 | 69,191 | 1.18 |
| 16. | Odisha | 672 | 77,150 | 0.87 |
| 17. | Kerala | 561 | 40,692 | 1.38 |
| Total | 81,073 | 19,83,178 | 4.09 |
The above shows that the PTR ranges very widely in India, from 0.87 in Odisha, 1.18 in Haryana, and 1.38 in Kerala to very high levels in Telangana (7.33) Maharashtra (11.45), Gujarat (7.69), Delhi (7.07). But, this also hides the fact that total number of cases could be kept low by merely not conducting enough tests. The case of Kerala is interesting especially as it has been touted by several commentators and the media as among the best examples of good pandemic management, even being compared with other sovereign nations.
Out of the 17 States, only Telangana has done less number of tests than Kerala. Even then, Odisha, Haryana, and Andhra Pradesh, all with higher number of tests than Kerala, have shown a lower PTR. Thus, the less number of tests is one big reasons why the numbers are less in Kerala. Further, leave aside the mysterious absence of the Tablighi effect on Kerala, and certain anecdotal accounts of positive cases emerging even a month after return from abroad, and that of death occurring a few days after returning from hospital, the low level of tests, in a State that has among the best medical infrastructure and personnel in the country, raises doubts as to whether Kerala is a victim of Goodhart’s Law.
Let us go back to a paper published in the Economic and Political Weekly (Vol. 34, No. 12, Mar. 20-26, 1999, pp. 713-716). Irudaya Rajan and Mohanachandran in their paper, “Estimating Infant Mortality in Kerala”, examine the validity of various mortality rates in Kerala. The performance in such rates are at the core of other human development indices, the basis for Kerala being hailed as a “model”. Reported Infant Mortality Rates (IMR) in Kerala had declined dramatically from 57 per thousand live births during 1971-76 to 43 during 1977-80, 32 during 1980-85, 24 during 1986-90, and further to 15 during 1991-96. It was 14 in 1996. Some of the year-to-year declines in IMR have been quite dramatic, too good to be true. For instance, a nine-point decline from 56 in 1976 to 47 in 1977 and again 28 in 1988 to 21 in 1989 and further to 15 in 1990. Today, the IMR stands at 7, down from 10 in 2016. These reductions were quite dramatic and often just 1/3rd of the second best.
But, the period also saw dramatic increase in still-births drawing Rajan and Mohanachandran to conclude that as “much emphasis is given to preventing infant mortality, some deaths which occurred immediately after the birth [were being] classified as ‘stillbirth’ by the medical practitioners deliberately”. If these deaths are counted as still births, they do not enter the IMR figures. Based on their investigations, they refixed the IMR at 37 instead of 14. These estimates, according to them “are well supported by the available indicators of Kerala: effective reproductive span of five years, 27 per cent low birth weight babies, stillbirth rate of 10 and one-week infant mortality of 11 per thousand live births.” This pioneering study has unfortunately not been followed up with further investigations, say by drilling down into district level and further disaggregated data to narrow down to the origin of the problem, if any.
This is not belittle the grand achievements of Kerala in the sphere of social development. This “model” is a story which has been in the making for over 200 years ever since the Vaccination Department was started in 1815. Panikkar (1975) narrates how the Travancore and Cochin States showed attention to the hygiene of public places like markets and cart-stands, which were brought under the Public Health Department in late 19th century, with a Market Inspector appointed around 1908. Child Welfare and Maternity Centres were set up in different places in 1933. Panikkar cites this decentralised and equitable distribution of healthcare facilities, the parallel development of western and indigenous systems of medicine, and their high level of usage, as responsible for better mortality rates and life expectancy even before independence. Behavioural hygiene practices influenced and facilitated by forty-odd rivers, and the prevalence of Ayurveda-based hygienic practices, even in day to day life, would also have contributed.
Coming back to Covid-19 and Goodhart’s Law, what questions do they leave us with. Even if you are best, the Goodhart’s Law comes into operation, in various ways, when measurable targets are put in place. Including for human development indices such as IMR and literacy rates. For instance, for the 1931 census in Travancore, the enumerator was required to assess literacy according to whether the respondent was able to write a letter to a friend, and whether he could read and understand his friend’s reply. Some decades later, the test got diluted to whether the respondent could write her name in her mother tongue. There is a need to continuously redefine and recalibrate the goal posts. There is also need for credible social audit outside the control of the government. Only then we will know whether we are Goodhart-ed.
