Convenient classification, collateral damage
The Supreme Court ban on alcohol sales near highways has people going to great lengths. Literally. One bar in Kerala built a maze to push the travel distance between the highway and the bar beyond the prescribed 500m dry zone.
The more common response, however, is simply to relabel the roads to make highways disappear and local roads emerge overnight. A 9 April Hindu BusinessLine article reported that “hundreds of kilometres” of highways in Maharashtra, Himachal Pradesh, Uttarakhand, Rajasthan, Punjab and West Bengal have recently become local, municipal or district roads. More states seem poised to follow the lead: the Aam Aadmi Party explicitly asked the Goa government to denotify state highways wherever legally possible in order to protect the tourist industry.
This is all well and good—more jugaad at work—but what happens when the next hapless researcher wants to investigate, say, “The dynamics of road development under competitive federalism”? The proportion of highway kilometres among roads is a (crude) summary statistic for the state of the road network. It will immediately look worse for FY 2017-18. The strange bulge in the share of local roads in 2017 will probably also be concentrated in areas with rapid economic change—the kinds of areas along the edges of cities and corridors where entrepreneurs are setting up brewpubs, hotels and restaurants. Who will be the first to spin a story about the links between regional growth or urbanization and the rise of local control as proxied by jurisdiction over infrastructure? Someday, somewhere, perhaps buried in comparative analysis, it will happen. Closer home, it’s a matter of time before a state or local politician claims credit for an amazing increase in the share of local roads under his or her watch.
The quality of India’s economic, environmental and social data has come under increasing scrutiny as researchers compare it to a wider array of alternative unofficial sources. Long-standing claims about methodological and process shortcomings are gaining more “oomph” as anomalies in the statistical picture of the country have emerged. How much water does India have? The answers from its scientists and its ministries vary. Weather? Skymet and the India Meteorological Department offer different pictures. Consumption? Take your pick—the National Sample Survey Office or Central Statistics Office or private market research. Growth? There is more variation between estimates than can be explained by different calculation methods alone. And luminosity data offer yet another picture.
The ongoing critique of data focuses on collection techniques and technical concerns rather than accusations of blatant politicization (I am going to leave controversies over definitions, such as those around urbanization or poverty, aside for now). People may argue that the government is picking its headline numbers politically, but few would contend that it is systematically manipulating the deeper collection processes behind them.
It doesn’t have to directly manipulate; policy frameworks that create incentives for convenient classification can have the same effect. Classifying for convenience is not conventionally “political” in the sense of manipulation of the big picture to gain points with voters or investors; but it is purposeful, and for purposes other than accurate descriptions of the state of the economy, polity and environment. Collateral damage is still damage.
Convenient classification is common. Forests, for example, are notified by the state without a uniform definition. And a designation as “forest” has implications for terms of industrial use as well as tribal rights. It is not surprising that some forests, especially those with mining and industrial development potential, don’t show up in the data. It’s not clear what effect this has on the overall forest area count, since there are other incentives to promote forests embedded in fiscal transfer rules. But it would be surprising if these pushes and pulls did not affect official data.
Similarly, every tax exemption—for research and development, for agricultural income, for service exports—not only encourages differences in real behaviour, but also incentivizes reclassification that distorts the statistical picture. It’s the same with tariff variations, differential regulation, or any other kind of targeted treatment for activities or entities that cannot be readily, cheaply, unambiguously audited. Ray Fisman and Shang-Jin Wei, for example, have a series of papers that show a correlation between the differences in tariff treatment for similar goods and mismatches between reported exports from one country and imports to a partner country.
Misreporting of trade in certain kinds of cars, cultural artefacts, or frozen poultry may not seem to matter much on the scale of things, but consider how the fiction might affect reports on an industry outlook or the analysis of sectoral transitions. A forensic accounting of the impact of industrial policy on data and our understanding of economic history could be fascinating, if uncomfortable.
Coming back, the pressures for more convenient classification are likely to build. Rules-based policy places more weight on metrics as gateways for valuable rights and concessions. The tightening of the tax net increases the impact of classification decisions, and thus the incentives to push them one way or another. Balancing the pressures for convenient classification, or at least recognizing that they exist when designing policy or allocating resources for enforcement, is an essential addition to the to-do list on data quality.
Jessica Seddon is managing director of Okapi Research and Advisory and visiting fellow at IDFC Institute. She writes fortnightly on patterns in public affairs.
Comments are welcome at firstname.lastname@example.org