Here’s the official note from the police regarding my stolen phone. If you see it, will you please let me know? And yes, they do ask you to put your father’s name on every official form here.
First look at methodology
Yesterday, I got a glimpse into Lokniti’s methodology for their state election surveys. They are quite thorough about their reporting. If you want to see an example for yourself, check out the methodological note they released for their 2009 National Election Study here (“The current NES is the largest and the most comprehensive social scientific survey of Lok Sabha Election…and perhaps of any election in the world.” Pretty cool.). I won’t summarize it in full just yet since I don’t have all the details. But the quick, dirty version for state election polls–like the one we are going to do in Bihar in the coming months–sounds something like this (this is not so different from their national surveys either; there is just an additional level of sampling at the beginning):
1) Probability-proportionate-to-size (PPS) sampling of assembly constituencies (ACs) is the first stage of sampling. This means that the Lokniti team randomly draws assembly constituencies to which they will send interviewers, with larger ACs more likely to be selected (here is a list of the assembly constituencies for National Capital Territory of Delhi, as an example). The population data they use for the ACs comes from the 2011 census (haven’t found exactly where this is online, but it should be somewhere around here on the Indian census website). Considerations related to costs and logistics means replacements of ACs sometimes have to be made.
2) Simple random sampling of four or five polling stations within the ACs is the next step (I think four or five. I will have to check again with Nitin about the exact number). Simple random sampling means that the Lokniti team does not weight the polling stations by population size when they sample them; it’s effectively the same as blindly pulling golfballs of the same weight and feel from a hat. Here are my home AC’s polling stations, for reference.
3) Simple random sampling of 35 registered voters within each polling station (again, I need to check with Nitin on the exact number). Lokniti expects a response rate of somewhere around 70% on average across polling stations such that their investigators (their word for interviewers) complete about 25 interviews per polling station. If selected respondents refuse or are not available, there can be replacement with someone else of the same gender and similar age from the same polling station.
What amazed me most about this procedure at first is that every registered voter’s name, address, and age is online and it is SO easily accessible. They even have maps of the neighborhoods so investigators can more easily track you down! Don’t believe me? See for yourself. Here’s one in Maharashtra. Here’s one in Tamil Nadu. Here’s one in Mizoram (amusingly, the Mizoram website adds the disclaimer, “To protect the privacy of electors, their photographs are not shown”). Here’s one in Kerala. Okay, you get the point.
It is incredibly easy to see who lives in what neighborhood, and who has registered to vote and who hasn’t. At first, this creeped me out. But after thinking about it, I realized that the United States is not so different. No, we don’t keep all that information online, but county and state boards of elections keep this information and do make it available. The Democratic and Republican parties pull these lists for voter contact sheets; it’s how that annoying Obama campaign volunteer knocking at your door knows the names and ages of registered voters in the house, and whether they voted in the last election.
Systematic sampling bias
So… thoughts about systematic bias in sampling here? Well, Lokniti is only talking to people who have registered to vote. If you’re doing an election survey and want to talk only to voters, this would certainly make the process more efficienct. Shreyas says that upwards of 90% of Indians are registered to vote even if not that many end up actually voting. I’m not sure if they use this method for other types of surveys; they certainly don’t for youth surveys. Lokniti will be doing at least one other type of national survey in the next year, so I’ll get to witness that firsthand and report back.
What about homeless people? There are a lot of homeless in India. If homeless people are registered to vote, and list somewhere in their AC as an address, they could theoretically be selected. If a permanent address must be listed to register to vote, then the homeless are effectively disenfranchised, and so they should be excluded from a survey of voters. I’ll have to look more into this issue as well.
Today, SPSS was installed on my computer and I received my first dataset. Yippee! Data! I will of course have to ask permission before posting any analysis here. But these datasets look incredibly rich and ripe for exploration.
My office is kind of a politics nerd’s dream. Everyone on the team is an avid follower of Indian politics, and also international politics, so we are always discussing the day’s events at work and at lunch. Today the big story was BJP’s victory in Bengaluru’s municipal elections. My office has two BJP supporters, and four whose views I won’t characterize beyond opposition to the BJP. It makes for interesting conversation and thoughtful debate; everyone is remarkably respectful of each other’s different views.
The date for the Bihar elections has not yet been announced. The election is set to happen in October, and parties are already angling for advantage. Last week Prime Minister Narendra Modi (of the BJP) announced a ridiculously large development package to entice voters (to non-Indian readers, 1 lakh=100 thousand, and 1 crore=10 million, so Modi’s announced package is 1.25 trillion rupees). And his opponents have now countered with… a different package, worth 19,500 crore rupees (195 billion rupees).
I will write an explainer on the Bihar elections once a date has been selected. It is both very complicated and quite simple. The basic narrative is that the BJP (Bharatiya Janata Party), a right-wing Hindu nationalist party, has been on a winning streak in Indian politics over the past few years. That streak was temporarily halted with the Delhi assembly elections in February of this year, in which the anti-corruption Aam Aadmi Party (AAP) won in a landslide. Was this a signal that BJP’s momentum is slowing at the national level? Or was it merely an idiosyncratic outcome of Delhi politics? Indians will be looking to Bihar to gauge the BJP’s strength going forward.