search

jessica-writes-code / MSE231-Assignment4

0 stars 0 forks source link

General Comments/Tracking #1

Open jessica-writes-code opened 7 years ago

jessica-writes-code commented 7 years ago

@poorna-kumar @binghailing

Earlier today we brainstormed questions that we could ask in our survey. I think that Poorna came up with a really great idea about AI/Universal Basic Income. Below, I tried to formalize the questions that we might ask: 1) Are you familiar with artificial intelligence? 2) Do you think that advancements in artificial intelligence might threaten your job? 3) Do you think that advancements in artificial intelligence might threaten the jobs of people that you know? 4) Would you support a policy that ensures all citizens an unconditional sum of money (i.e., "Universal Basic Income")?

In the last question, we might provide a link to the wiki for "universal basic income".

Please let me know if you have edits/suggestions or if you would prefer to consider other topics. The other primary one that was suggested was about how individuals get their health insurance and how they feel about the ACA.

jessica-writes-code commented 7 years ago

To Do:

poorna-kumar commented 7 years ago

Some relevant links (we might want to include them as references in the report -- don't waste your time clicking and reading; I'm just creating an archive here because I don't want to lose the links): http://link.springer.com/article/10.3758/s13428-013-0365-7#page-1 http://cdp.sagepub.com/content/23/3/184.full

jessica-writes-code commented 7 years ago

Other ideas:

1) Health insurance. Questions to include: a. Do you have health insurance? b. If so, how do you acquire your health insurance? (e.g., job, state-funded system, exchange, direct-purchase, etc.) c. How do you feel about the Affordable Care Act? (e.g., Very positive, Somewhat positive, etc.)

2) Women's health. Questions to include: a. How do you feel about the Planned Parenthood organization? b. Should insurance companies be required to cover birth control? c. Should individuals have access to abortion services?

3) Climate change. Questions to include: a. Do you believe that climate change is occurring? b. If so, do you believe that climate change is human-caused? c. How do you feel about President-elect Trump's proposal to quit the Paris Climate Agreement?

I literally cannot think of anything not political right now. I'm sorry about that.

poorna-kumar commented 7 years ago

One of my concerns about our old idea (and all of these) is that they're basically opinion surveys: not experiments as such. No harm in doing that, but if we could come up with an experiment to test a hypothesis that would be more in line with what we've been learning.

On Sun, Nov 13, 2016 at 7:02 PM, Jessica notifications@github.com wrote:

Other ideas:

1) Health insurance. Questions to include: a. Do you have health insurance? b. If so, how do you acquire your health insurance? (e.g., job, state-funded system, exchange, direct-purchase, etc.) c. How do you feel about the Affordable Care Act? (e.g., Very positive, Somewhat positive, etc.)

2) Women's health. Questions to include: a. How do you feel about the Planned Parenthood organization? b. Should insurance companies be required to cover birth control? c. Should individuals have access to abortion services?

3) Climate change. Questions to include: a. Do you believe that climate change is occurring? b. If so, do you believe that climate change is human-caused? c. How do you feel about President-elect Trump's proposal to quit the Paris Climate Agreement?

I literally cannot think of anything not political right now. I'm sorry about that.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-260238304, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_b3oWSWK7R9G0UKIvsi00OpDsBveAks5q989MgaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Ah. I figured that the hypothesis would essentially be that individuals who believe their jobs might be threatened (or that their friends' jobs might be threatened by AI) would be more inclined to support UBI.

Similarly, I think that the health insurance questions would serve to test the (related) hypotheses that people who currently use the exchanges like the ACA and, more significantly, that people who get their health insurance through their job do not like the ACA. We could even add a question about how people got their health insurance pre-ACA (e.g., in 2009) and evaluate the opinions of the people who switched.

If we want to test a hypothesis that's unrelated to people's opinions about things, I think we'll have to search a little harder. I'm having difficulty coming up with anything especially interesting off of the top of my head.

Only one major idea: If we want to go a little off-book, we can recreate one of those "James" and "Jane" tests, where you have people read a paragraph or two about someone and then answer questions about the person they read about. You randomly assign people to be reading about "James" or "Jane" but keep all the rest the same. This is an experiment that's been done plenty of times (in plenty of different contexts -- gender, race, etc.). The only "twist" on it that I could see being interesting would be randomizing people to be reading about a gay couple vs. a straight couple; within that you could even do gay men vs. gay women. Hesitations on that: 1) We might be toeing some social lines that it would be tricky to discuss in a straight-forward manner, 2) it's not exactly to the letter of the HW assignment, and 3) I have literally no idea how to randomize things w/in surveys on Amazon Mechanical Turk. (This blog discusses the latter: http://babieslearninglanguage.blogspot.com/2013/10/randomization-on-mechanical-turk_10.html)

jessica-writes-code commented 7 years ago

I'm sorry that you just sort of got my stream-of-conciousness ramblings...

poorna-kumar commented 7 years ago

I'm sorry -- I misread the assignment question. D'oh. 😌 I thought the natural experiment was supposed to be done with Turk, but they're unrelated. So any of the previous questions are good with me.

And the James/Jane test sounds really nice: might be a good thing to write about for our natural experiment. There's also another idea I had, which is to measure the effects of polls on voting patterns (it looks like a lot of work has already been done on this topic, which reduces its sheen somewhat).

On Sun, Nov 13, 2016 at 7:22 PM, Jessica notifications@github.com wrote:

I'm sorry that you just sort of got my stream-of-conciousness ramblings...

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-260240285, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_b6F97Kuo8kEF4DtZyHXayaM0eyNmks5q99P9gaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

That said, we now have the opportunity to do a pretty immediate post-election survey. So, that could be our bit of sheen.

Anyway, maybe let's briefly discuss after class and settle on one so that we can do a test run this evening.

On Sun, Nov 13, 2016 at 11:31 PM, poorna-kumar notifications@github.com wrote:

I'm sorry -- I misread the assignment question. D'oh. 😌 I thought the natural experiment was supposed to be done with Turk, but they're unrelated. So any of the previous questions are good with me.

And the James/Jane test sounds really nice: might be a good thing to write about for our natural experiment. There's also another idea I had, which is to measure the effects of polls on voting patterns (it looks like a lot of work has already been done on this topic, which reduces its sheen somewhat).

On Sun, Nov 13, 2016 at 7:22 PM, Jessica notifications@github.com wrote:

I'm sorry that you just sort of got my stream-of-conciousness ramblings...

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1# issuecomment-260240285, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_ b6F97Kuo8kEF4DtZyHXayaM0eyNmks5q99P9gaJpZM4Kvew1

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-260266258, or mute the thread https://github.com/notifications/unsubscribe-auth/AKKw-oahD-WKW_1e8kB2SnUGhSqpbjsRks5q-A5GgaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Okay, based on our discussion earlier today, we are going to go with the AI-related survey. Below is a proposal for the questions. These are (obviously) in addition to the basic demographic questions that we are required to ask by the assignment.

1) How would you describe your political affiliation?

2) Do you think that advancements in automation and/or artificial intelligence might eliminate your job (or primary source of income)?

3) Would you support a policy of "Universal Basic Income" (as described below)?

"Universal Basic Income" is a form of social security in which all citizens or residents of a country regularly receive an unconditional sum of money, either from a government or some other public institution, in addition to any income received from elsewhere. The amount from the government is enough to purchase basic necessities (e.g., food, water, housing).

For additional details, please see: https://en.wikipedia.org/wiki/Basic_income

@poorna-kumar - Can you please either edit or approve the above? Then, @binghailing can implement the trial.

poorna-kumar commented 7 years ago

Hi Jessica --

Two suggestions:

  1. Reduce the number of political affiliations from five to three? (Or maybe we should keep it at five, and then collapse it to three if we see fit later on).
  2. Maybe specify "advancements in AI will eliminate your job in the next 5/10/25 years"? A timeframe would be helpful for us to understand how immediate people perceive the threat of AI to be.

Poorna

On Mon, Nov 14, 2016 at 3:36 PM, Jessica notifications@github.com wrote:

Okay, based on our discussion earlier today, we are going to go with the AI-related survey. Below is a proposal for the questions. These are (obviously) in addition to the basic demographic questions that we are required to ask by the assignment.

1) How would you describe your political affiliation?

  • Very conservative
  • Somewhat conservative
  • Moderate
  • Somewhat liberal
  • Very liberal

2) Do you think that advancements in artificial intelligence might eliminate your job (or primary source of income)?

  • Yes
  • Unsure
  • No

3) Would you support a policy of "Universal Basic Income" (as described below)?

"Universal Basic Income" is a form of social security in which all citizens or residents of a country regularly receive an unconditional sum of money, either from a government or some other public institution, in addition to any income received from elsewhere. The amount from the government is enough to purchase basic necessities (e.g., food, water, housing).

For additional details, please see: https://en.wikipedia.org/wiki/ Basic_income

  • Yes
  • Unsure
  • No

@poorna-kumar https://github.com/poorna-kumar - Can you please either edit or approve the above? Then, @binghailing https://github.com/binghailing can implement the trial.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-260499400, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_byL__SlXwX6IkXAT62ZhSuGczrwjks5q-PB_gaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Great. Thank you, Poorna! Revised version below:

1) How would you describe your political affiliation?

2) Do you think that advancements in automation and/or artificial intelligence might eliminate your job in the next 15 years (or primary source of income)?

3) Would you support a policy of "Universal Basic Income" (as described below)?

"Universal Basic Income" is a form of social security in which all citizens or residents of a country regularly receive an unconditional sum of money, either from a government or some other public institution, in addition to any income received from elsewhere. The amount from the government is enough to purchase basic necessities (e.g., food, water, housing).

For additional details, please see: https://en.wikipedia.org/wiki/Basic_income

Look good to you?

poorna-kumar commented 7 years ago

Yes. Looks great! I'm only wondering if my suggestion of collapsing the number of groups was a good idea. :P I guess I leave the final call on that to you both.

One more small change: "job (or primary source of income) in the next 15 years".

Btw, thanks a lot for keeping us on track, Jessica! I'd never have started so early, but I'm so grateful we're doing so.

On Nov 14, 2016 6:36 PM, "Jessica" notifications@github.com wrote:

Great. Thank you, Poorna! Revised version below:

1) How would you describe your political affiliation?

  • Conservative
  • Moderate
  • Liberal

2) Do you think that advancements in automation and/or artificial intelligence might eliminate your job in the next 15 years (or primary source of income)?

  • Yes
  • Unsure
  • No

3) Would you support a policy of "Universal Basic Income" (as described below)?

"Universal Basic Income" is a form of social security in which all citizens or residents of a country regularly receive an unconditional sum of money, either from a government or some other public institution, in addition to any income received from elsewhere. The amount from the government is enough to purchase basic necessities (e.g., food, water, housing).

For additional details, please see: https://en.wikipedia.org/wiki/ Basic_income

  • Yes
  • Unsure
  • No

Look good to you?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-260530330, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_byRkVVecU9qEGssz6udNJoq5g3hgks5q-RqogaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

I think collapsing is probably a good idea. The additional layers just introduce variance without likely giving us much else to work with (since we are only surveying 200 people).

I fixed the grammatical error you found. So, final draft is below. @binghailing - Can you start the survey test run?

1) How would you describe your political affiliation?

2) Do you think that advancements in automation and/or artificial intelligence might eliminate your job (or primary source of income) in the next 15 years?

3) Would you support a policy of "Universal Basic Income" (as described below)?

"Universal Basic Income" is a form of social security in which all citizens or residents of a country regularly receive an unconditional sum of money, either from a government or some other public institution, in addition to any income received from elsewhere. The amount from the government is enough to purchase basic necessities (e.g., food, water, housing).

For additional details, please see: https://en.wikipedia.org/wiki/Basic_income

poorna-kumar commented 7 years ago

http://economicspsychologypolicy.blogspot.com/2015/06/list-of-19-natural-experiments.html -- some more inspiration for part 2 of our HW.

jessica-writes-code commented 7 years ago

Hi all -

We've floated some ideas about the natural experiment that we could discuss. I wanted to get those down and add a few more. See below:

1) Bag taxes.

Some cities that border each other have different bag tax policies. We could investigate how this affects consumer spending in those cities, likelihood of shopping patterns changing after the tax is implemented, etc.

2) Redistricting and political engagement.

There seems to be some literature on how winner-take-all systems, like the Electoral College, tend to depress voter turnout, because people don't think their vote will matter. People who vote in line with the expected majority don't bother because the region will swing their way in spite of their inaction and people who vote against the expected majority don't bother because they know they'll be out-voted. (Empirically, it seems that there is higher voter turnout in battleground states: http://qcpages.qc.edu/political_science/profmat/Lipsitz_Consequences%20of%20Battleground.pdf, http://www.usatoday.com/story/news/politics/2012/12/23/voter-turnout-swing-states/1787693/)

Every 10 years (with the census), Congressional representative districts are redrawn. We could analyze whether voter participation (per capita) decreases when an area moves from a relatively competitive district to a less competitive district (or vice-versa).

3) Mariel Boatlift and xenophobia

Per the article Poorna linked, in 1980 a significant number of Cubans entered the US and took up residence in Miami. One could look at public opinions, campaign speeches, etc. to see if Miami became more or less xenophobic over the course of the next few years. This would be a differences-in-differences analysis -- we'd need to compare pre/post Mariel Boatlift Miami to pre/post Mariel Boatlift other cities.

How do we feel about any of the above? Any other ideas?

Also, the bit about "16 and pregnant" in the article Poorna linked is hilarious.

jessica-writes-code commented 7 years ago

@poorna-kumar @binghailing

Hi all -

I wrote up a draft of section 2 of part I and started writing some things for part II. I have one concern about our idea for part II. The assignment says:

You do not need to carry out any data analysis, but the natural experiment you describe should be based on a specific instance of as-if randomization that does in fact exist in the real-world. That is, it should be possible in theory to carry out your proposal.

However, because California banned plastic bags in 2016, we couldn't actually currently carry out our proposal (since the natural experiment no longer exists). Do the two of you think that our idea violates his instructions or am I reading too much into the phrasing?

Jessica

jessica-writes-code commented 7 years ago

So, I've done slightly more thinking about the above. I think our idea is still okay. While the randomization does not currently exist, it did at one point. Given how difficult it is to find a reasonable natural experiment, I speculate that is sufficient. Moreover, we could still carry out the analysis; we'd just be relying on historical data. And, realistically, that's the case for all major papers on natural experiments -- e.g., no one was writing about the draft as an experiment at the time of the draft; all those papers were written later.

Please speak up if you disagree and think we should change the idea.

poorna-kumar commented 7 years ago

I agree with this.

On Nov 22, 2016 8:41 AM, "Jessica" notifications@github.com wrote:

So, I've done slightly more thinking about the above. I think our idea is still okay. While the randomization does not currently exist, it did at one point. Given how difficult it is to find a reasonable natural experiment, I speculate that is sufficient. Moreover, we could still carry out the analysis; we'd just be relying on historical data. And, realistically, that's the case for all major papers on natural experiments -- e.g., no one was writing about the draft as an experiment at the time of the draft; all those papers were written later.

Please speak up if you disagree and think we should change the idea.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262293425, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_b8j6ubg6KdwDOFpfoWbOW2g3-dsZks5rAxs0gaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Great. Thank you, Poorna.

I started writing some stuff about it; I'm having a little difficulty figuring out if I've said what needs to be said in the methodology section. So if someone would like to take a look at that and add some stuff, it would be great.

I also haven't done anything to the caveats section, largely because I'm having difficulty thinking about major issues. The ones that strike me initially are:

However, both of those seem pretty obvious, so I'm searching for something more...

poorna-kumar commented 7 years ago

Thanks for your note, Jessica. I looked at the document, and what's been written so far is great. That said, I think we can add a few things to the document:

  1. We should probably define exactly what is our response variable. I propose average (median?) daily sales in dollars, but this is pretty arbitrary and I haven't thought it through. Please suggest alternative response variables if this seems flawed.

  2. To build off point 1, suppose we look at two adjacent areas, one with a bag ban (Area 1) and one without (Area 2). Then, we should be able to identify businesses of interest in each area and put them into categories (small/medium/large and/or grocery/pharmacy/clothing/etc.). After that, for each store we have, we look at the change in sales before and after the treatment began, i.e., we compute, for store i, we compute metric(i) = average_daily_sales_before(i) - average_daily_sales_after(i). We need to define how to compute this average: I assume we will identify a period of interest (how long?) and take the average across that period. Then, we can compute this metric for every store i, and find the average value of this metric in each category j of store in Area 1 and Area 2. So, for each category j of store, we can find the difference in this metric between Area 1 and Area 2, and test whether this is significantly different from zero.

  1. We could write a little bit about how we bucket businesses -- do we want to bucket them based on products (grocery stores, pharmacies, clothing, etc.) or size (small, medium, large) or both? The more groups we have, the less data we have per group, but the more we are controlling for.

I've edited the document a little bit, and will edit it some more in the next hour or so. I've also left some comments in the document. I guess this is one of those assignments where the more we think about our problem, the more choices we realize we could make, and the more choices we make, the more caveats and limitations we impose. :P

poorna-kumar commented 7 years ago

I got a lot of ideas from reading the beginning of the Card paper, but I haven't finished reading it. I'll go over it a bit more tonight and make some more changes.

jessica-writes-code commented 7 years ago

Thanks so much for your reply, Poorna! On the issues that you pointed out:

1/2. I agree. I think defining a metric is important and I think "average daily profit" is appropriate. (See response to your first comment in the document for the justification for "profit" over "sales" or "revenue", though ultimately not a huge deal.)

Also, agreed with the computation. As far as time-horizon, I suggest computing the avg_daily_profit_before as sales during the 4-week period that end a month before the bag ban and then avg_daily_profit_after as sales during the 4-week period that begins a month after the bag ban. (I recognize that is a bit difficult to phrase...) This addresses a few issues:

I also agree with your assessment of how to generalize the method. Calculating for every pair of areas seems like the most effective procedure. Should there be places (and likely, there are) in which a treated area is next to two un-treated areas, I would agree that we not use those areas in our analysis. They add needless complication.

  1. Excellent. Thank you! I've added a bit about the bucketing. I've suggested grocery, pharmacy, and department stores, as these seem most likely to be impacted. I've also suggested division into small, medium, and large. I suspect that if we look at the entirety of LA County, we should not have a huge problem with not having enough data. (Though, I don't really know the area, so maybe @binghailing can comment?)

Generally, I've responded to your comments in the document and added a few of my own. LMK if I've missed any.

poorna-kumar commented 7 years ago

I've added a bit more to the document, and listed some caveats as comments.

I think we're at a good length for methodology/caveats now (although of course it might need to be edited and/or cut down a bit). Maybe we should add a bit more about the social science question we're asking (for example, why it's interesting, theories for what could happen to businesses after a bag ban, etc.).

poorna-kumar commented 7 years ago

I was just thinking about our idea, and I realized, if we focus on areas just near the border, won't we be giving customers the option to go to another store across the border, which doesn't have the bag ban?

So we are not really measuring the impact of the bag ban on businesses, but instead the impact of a piecemeal bag ban on businesses which are close to competing establishments which do not have the ban. I wonder if this changes things a lot.

On Wed, Nov 23, 2016 at 6:33 PM, Jessica notifications@github.com wrote:

Thanks so much for your reply, Poorna! On the issues that you pointed out:

1/2. I agree. I think defining a metric is important and I think "average daily profit" is appropriate. (See response to your first comment in the document for the justification for "profit" over "sales" or "revenue", though ultimately not a huge deal.)

Also, agreed with the computation. As far as time-horizon, I suggest computing the avg_daily_profit_before as sales during the 4-week period that end a month before the bag ban and then avg_daily_profit_after as sales during the 4-week period that begins a month after the bag ban. (I recognize that is a bit difficult to phrase...) This addresses a few issues:

  • reactions to the change: people might start preparing for the bag ban, when they know it's coming. I'm not sure if/how that would affect sales, but it's probably smart to avoid capturing data during the time period because of it. More significantly, for some time period after the ban, people either might not be aware of it or might not have adjusted to it (e.g., they forget, and go shopping when they otherwise wouldn't). The month delay before we start sales calculations means we don't erroneously capture those effects.
  • intra-week flux in profit: if we choose a timespan defined by a month or period of days, we stand some risk of capturing more weekends (or whatever constitutes a high-spending day) in the before time period than the after time period (or vice versa). This would artificially bias our results.

I also agree with your assessment of how to generalize the method. Calculating for every pair of areas seems like the most effective procedure. Should there be places (and likely, there are) in which a treated area is next to two un-treated areas, I would agree that we not use those areas in our analysis. They add needless complication.

  1. Excellent. Thank you! I've added a bit about the bucketing. I've suggested grocery, pharmacy, and department stores, as these seem most likely to be impacted. I've also suggested division into small, medium, and large. I suspect that if we look at the entirety of LA County, we should not have a huge problem with not having enough data. (Though, I don't really know the area, so maybe @binghailing https://github.com/binghailing can comment?)

Generally, I've responded to your comments in the document and added a few of my own. LMK if I've missed any.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262676753, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_b0muZfGxxxN6ElzaWHs7mOgpEaGYks5rBPeJgaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

So, I'd had the same thought. And, I agree that we're not necessarily going to capture the same effect as if bags were banned in their entirety. We should certainly include this issue in the caveats section.

I do still think that the information we are capturing is important. It certainly answers questions about consumer behavior changes in response to the bag ban. If we see non-bag-ban stores have revenue increases that are roughly proportional to the profit decreases in bag-ban stores, we can say that consumers do respond. And they respond in ways that hurt stores subject to the bag ban. The social science implication is then that, when implementing a bag ban, it needs to be done in a large-scale way, to prevent disadvantaging the businesses that happen to be subject to it. So, that isn't exactly "bag-ban is good" or "bag-ban is bad", but it does have serious policy implications.

Also, if we want to just test the expense issue, maybe we could suggest capturing expense as a proportion of revenue? That would tell us if bag-ban businesses have their expenses per sale go up vs. non bag-ban businesses.

Does the above seem logical?

On Thu, Nov 24, 2016 at 10:34 AM, poorna-kumar notifications@github.com wrote:

I was just thinking about our idea, and I realized, if we focus on areas just near the border, won't we be giving customers the option to go to another store across the border, which doesn't have the bag ban?

So we are not really measuring the impact of the bag ban on businesses, but instead the impact of a piecemeal bag ban on businesses which are close to competing establishments which do not have the ban. I wonder if this changes things a lot.

On Wed, Nov 23, 2016 at 6:33 PM, Jessica notifications@github.com wrote:

Thanks so much for your reply, Poorna! On the issues that you pointed out:

1/2. I agree. I think defining a metric is important and I think "average daily profit" is appropriate. (See response to your first comment in the document for the justification for "profit" over "sales" or "revenue", though ultimately not a huge deal.)

Also, agreed with the computation. As far as time-horizon, I suggest computing the avg_daily_profit_before as sales during the 4-week period that end a month before the bag ban and then avg_daily_profit_after as sales during the 4-week period that begins a month after the bag ban. (I recognize that is a bit difficult to phrase...) This addresses a few issues:

  • reactions to the change: people might start preparing for the bag ban, when they know it's coming. I'm not sure if/how that would affect sales, but it's probably smart to avoid capturing data during the time period because of it. More significantly, for some time period after the ban, people either might not be aware of it or might not have adjusted to it (e.g., they forget, and go shopping when they otherwise wouldn't). The month delay before we start sales calculations means we don't erroneously capture those effects.
  • intra-week flux in profit: if we choose a timespan defined by a month or period of days, we stand some risk of capturing more weekends (or whatever constitutes a high-spending day) in the before time period than the after time period (or vice versa). This would artificially bias our results.

I also agree with your assessment of how to generalize the method. Calculating for every pair of areas seems like the most effective procedure. Should there be places (and likely, there are) in which a treated area is next to two un-treated areas, I would agree that we not use those areas in our analysis. They add needless complication.

  1. Excellent. Thank you! I've added a bit about the bucketing. I've suggested grocery, pharmacy, and department stores, as these seem most likely to be impacted. I've also suggested division into small, medium, and large. I suspect that if we look at the entirety of LA County, we should not have a huge problem with not having enough data. (Though, I don't really know the area, so maybe @binghailing https://github.com/binghailing can comment?)

Generally, I've responded to your comments in the document and added a few of my own. LMK if I've missed any.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1# issuecomment-262676753, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_ b0muZfGxxxN6ElzaWHs7mOgpEaGYks5rBPeJgaJpZM4Kvew1

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262829877, or mute the thread https://github.com/notifications/unsubscribe-auth/AKKw-ghXOBoUjkF5FaovsljluCd_ojDUks5rBdiYgaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Also, I think we might be able to reasonably make the assumption that the bag ban is annoying to consumers, but really only has effects at the margin -- i.e., they choose to walk into a store to pick up something small or not. I don't think that it would be so impactful as to cause them to drive out of their way to go shopping elsewhere (the gas costs would off-set the gain from not paying for bags).

poorna-kumar commented 7 years ago

Thanks, Jessica. You're right that we are still asking an important question -- if we do see a change in profit before and after the ban for stores that are affected by the ban, we can conclude that either consumers are going elsewhere, or it just organically hurts business profits because customers don't like it: either is a significant outcome.

I am not sure if there is any way to tell whether a store's profits fell because: a. customers were put off by the bag ban and just shopped less, or b. whether customers moved to a competing establishment because they could. So I'm not sure how to disentangle these two conclusions, and draw a single policy conclusion from it. For example, what if customers shopped less overall (because the bag ban was an inconvenience), but also chose to shop a larger percent of the time at establishments without the ban? If we are comparing two establishments that are very close, wouldn't it be very easy for a customer in a car to choose which one to go to? If it's just a few hundred meters of distance, the gas charges would be quite low too, right? I guess in the absence of tracking information for individual customers, we can't really say which of the two happened.

I had a look at the Card paper again, and they had recorded that food prices in NJ increased after the increase in minimum wage. So there was an incentive for customers to eat in Pennsylvania rather than in NJ, and it's possible in theory that employment in NJ could have fallen as a result of customers' eating elsewhere and establishments having to downsize in response. However, since they recorded that employment didn't decrease, this was not an important effect to consider, as it wouldn't have changed their conclusion (admittedly, I didn't read the paper till the end).

I suppose we can say that we'd do the same as them: ignore this fact if we find that profits don't decrease. But if they do, then we can say that one of two things occurred.

jessica-writes-code commented 7 years ago

Okay. Having thought about this for another 15 minutes...

My suggestion: We change our metric to be expenses divided by revenue (i.e., expenses per sale) and compare that between stores on each side of the border. This answers the question of whether a business's expenses go up nontrivially, which is an interesting social science question about the implementation of bag bans, more generally (not just in implementation). It also controls for the issue of sales movement across the border, because it's a ratio, rather than a total quantity of sales.

Alternatively, we can do as you suggest. That method is just a bit harder to explain, is all.

poorna-kumar commented 7 years ago

I'm not sure what expenses as a proportion of revenue would capture: I would expect it would change in fairly predictable ways with a bag tax, right? (More expenses per sale because the store pays a tax on the bag it sold, and more revenue per sale because the customer pays for the bag too?) And ultimately, it's not a good measure of robustness of business -- even if a store had low expenses to revenue per sale, it could still be doing very badly because of fewer transactions.

My inclination is to go with what we had earlier. I was heartened by the fact that the Card paper also basically ignored this effect. We could include this in the caveats and limitations section.

jessica-writes-code commented 7 years ago

Alrighty. Sounds good to me.

On Thu, Nov 24, 2016 at 11:49 AM, poorna-kumar notifications@github.com wrote:

I'm not sure what expenses as a proportion of revenue would capture: I would expect it would change in fairly predictable ways with a bag tax, right? (More expenses per sale because the store pays a tax on the bag it sold, and more revenue per sale because the customer pays for the bag too?) And ultimately, it's not a good measure of robustness of business -- even if a store had low expenses to revenue per sale, it could still be doing very badly because of fewer transactions.

My inclination is to go with what we had earlier. I was heartened by the fact that the Card paper also basically ignored this effect. We could include this in the caveats and limitations section.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262837331, or mute the thread https://github.com/notifications/unsubscribe-auth/AKKw-g3c3zBtTfbK-wX1zQEDHqRumMj8ks5rBepQgaJpZM4Kvew1 .

poorna-kumar commented 7 years ago

Silly question: could you point me to a source that describes the plastic bag ban and tax in LA? When I searched, I could only seem to find a plastic bag ban, along with a ten-cent fee on paper bags.

jessica-writes-code commented 7 years ago

Hi Poorna, I don't think I quite understand the question.

I agree that LA only implemented a plastic bag ban & 10-cent fee on paper bags. Before that, I believe, a number of areas scattered throughout LA county had their own bag bans/taxes. I thought we were studying the latter. I actually had difficulty finding anything that discussed those taxes, specifically, but I'd gotten that understanding from @binghailing . However, we can certainly remove the discussion of taxes, since that will likely make the entire write-up cleaner.

poorna-kumar commented 7 years ago

Hmm, so what I could find said that there were a lot of regions in LA that had implemented a bag ban plus fee for paper bags. I couldn't find anything about another tax. I guess we can just talk about one treatment and one control then. (In fact, it would make our write-up simpler, which would be good).

On Nov 24, 2016 12:31 PM, "Jessica" notifications@github.com wrote:

Hi Poorna, I don't think I quite understand the question.

I agree that LA only implemented a plastic bag ban & 10-cent fee on paper bags. Before that, I believe, a number of areas scattered throughout LA county had their own bag bans/taxes. I actually had difficulty finding anything that discussed those taxes, specifically, but I'd gotten that understanding from @binghailing https://github.com/binghailing . However, we can certainly remove the discussion of taxes, since that will likely make the entire write-up cleaner.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262841494, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_bwQ_brN-M61R9mgq4O_GOkNfItVfks5rBfQ_gaJpZM4Kvew1 .

jessica-writes-code commented 7 years ago

Yeah. And having to say "bag ban and/or tax" every time was getting tedious.

Just discussing the ban/fee sounds good to me

On Thu, Nov 24, 2016 at 12:40 PM, poorna-kumar notifications@github.com wrote:

Hmm, so what I could find said that there were a lot of regions in LA that had implemented a bag ban plus fee for paper bags. I couldn't find anything about another tax. I guess we can just talk about one treatment and one control then. (In fact, it would make our write-up simpler, which would be good).

On Nov 24, 2016 12:31 PM, "Jessica" notifications@github.com wrote:

Hi Poorna, I don't think I quite understand the question.

I agree that LA only implemented a plastic bag ban & 10-cent fee on paper bags. Before that, I believe, a number of areas scattered throughout LA county had their own bag bans/taxes. I actually had difficulty finding anything that discussed those taxes, specifically, but I'd gotten that understanding from @binghailing https://github.com/binghailing . However, we can certainly remove the discussion of taxes, since that will likely make the entire write-up cleaner.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1# issuecomment-262841494, or mute the thread https://github.com/notifications/unsubscribe- auth/ASd_bwQ_brN-M61R9mgq4O_GOkNfItVfks5rBfQ_gaJpZM4Kvew1 .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-262842235, or mute the thread https://github.com/notifications/unsubscribe-auth/AKKw-vit4bLe4wLZeGp25opitqmcB2Z4ks5rBfZAgaJpZM4Kvew1 .

poorna-kumar commented 7 years ago

Discovered today: an interesting regression discontinuity at http://voxeu.org/article/cheaper-flights-and-scientific-collaboration.

jessica-writes-code commented 7 years ago

Nifty. Thanks for sharing!

jessica-writes-code commented 7 years ago

I've done some more editing to the report. Specifically, ...

Also, I left a few comments regarding sentences that I commented-out.

jessica-writes-code commented 7 years ago

@poorna-kumar @binghailing

Hi all -

I noticed this morning that Part I had still not been written. I've filled it out, so that it is now at least in a submittable form (i.e., no notes, etc.). Please feel free to revise or rewrite, but I wanted to make sure we had something that we could turn in this evening.

Jessica

poorna-kumar commented 7 years ago

Hi Jessica,

Sorry about that! I didn't end up having the energy to do it last night (my bad), so I was planning to get to it today morning. It looks like you've done a very good job on it -- I'm reading through it and trying to identify what I can add to it. It should be in submittable form by class time today.

Poorna

On Wed, Nov 30, 2016 at 7:23 AM, Jessica notifications@github.com wrote:

@poorna-kumar https://github.com/poorna-kumar @binghailing https://github.com/binghailing

Hi all -

I noticed this morning that Part I had still not been written. I've filled it out, so that it is now at least in a submittable form (i.e., no notes, etc.). Please feel free to revise or rewrite, but I wanted to make sure we had something that we could turn in this evening.

Jessica

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/jmoore523/MSE231-Assignment4/issues/1#issuecomment-263900870, or mute the thread https://github.com/notifications/unsubscribe-auth/ASd_b0djdzAp8kdc9RkOX1evmJVVK8Rsks5rDZT9gaJpZM4Kvew1 .