5 ms·

Would you mind quickly summarizing what that plot is telling us? I looked at it but it looks like I've got to read more of the paper before I understand it.

pge 7y ago
The plot is showing that there is a negative selection bias in the reporting of stories about minimum wage. That is, that stories suggesting that increasing minimum wage has negative effects (increases unemployment, etc) are overreported relative to stories showing a positive effect from an increase in minimum wage. I am only repeating what the paper says; I don't claim to have validated their methodology or sources.
- apcragg 7y ago
  That is an incomplete summary. Their conclusion is that despite this larger than normal selection bias, the minimum wage appears to have no impact on unemployment. The paper's abstract, discussion, and summary all mention this as part of the central thesis - why leave it out?
  - gowld 7y ago
    It's not an HN commenter's obligation to post everything a reader might want to see, certainly not highly technical point about the distinction between "this paper claims that evidence for an employment effect is over-reported" and "this paper claims that employment effect is approximately 0". They linked to the whole paper for the curious, and answered a particular question about it. Also, the use of "despite" does not logically fit the contents of (my) grandparent poster and parent poster, as grandparent poster didn't claim mention employment as a possible positive effect of minimum wage, and even within context of my parent post, "beyond" is a better word choice that n "despite"
  - scarejunba 7y ago
    I think summaries (by their nature) must leave something out. Considering the abstract mentions what you said (and that it's reasonable for me to have read the abstract), I think it was reasonable for the responder to focus their attention on the plot alone (which was the focus of my question).
  - IanCal 7y ago
    It is an incomplete summary of the paper, however it seems to be a pretty solid summary of what the plot is telling us.
  - roenxi 7y ago
    The obvious starting position is that raising minimum wages has a negative effect on the number of low wage employees hired; because the money from higher wages has to come from somewhere and might not be available - leading to reduced business activity. This is also the classic economics position. With that in mind, there is going to be a bias towards publishing the 'surprising' results that there is no impact/a positive impact from minimum wages. The bias that this paper is identifying is surely a real thing, but it might potentially be a bais against, eg, poorly run studies that pick up spurious effects. So the conclusions the paper draws w.r.t. elasticity aren't that interesting; we'd need to know why the bias exists before it is particularly useful. I'm a bit suspicious of using 'meta-analysis' at all because that just means that the faction that repeats themselves the most with low-quality research gets to pick what the elasticity is. In theory we shouldn't need 200 studies to make a claim, we should need 1 or maybe a small suite (like, 5) of very high quality studies that intellectually honest economists struggle to rebut.
  - pingyong 7y ago
    >why leave it out? Because he summarized the plot, not the paper.
- scarejunba 7y ago
  Much appreciated. And thank you to killjoywashere for selecting out the part of the paper that answers my question; gowld for reformatting appropriate to my mobile device. Apologies for not replying directly. I have a commenting rate-limiter applied to my account.
  - Fjolsvith 7y ago
    I, too, have a commenting rate-limiter applied to my account. I can only post about 5 comments in a day before HN tells me to "Slow Down, You're Posting Too Fast".
killjoywashere 7y ago
From the paper: ‘The simplest and most commonly used method to detect publication selection is an informal examination of a funnel plot’ (Sutton et al. 2000a: 1574). A funnel graph is a scatter diagram of precision versus estimated effect (such as estimated elasticities, regression coefficients or partial correlation coefficients). Precision is best measured by the inverse of the standard error (1/Se). As the name suggests, the expected shape is an inverted funnel — in the absence of publication selection. When there is no publication selection, estimates should vary randomly and symmetrically around the ‘true’ population effect. Because small-sample studies with typically less precision form the base of the graph, the plot will be more spread out there than at its top. However, it is the graph’s symmetry (or its absence) that is crucial for assessing publication selection (see Figure 1).
- michaelmrose 7y ago
  This method of quoting is unreadable on mobile
  - gowld 7y ago
    >‘The simplest and most commonly used method to detect publication selection is an informal examination of a funnel plot’ (Sutton et al. 2000a: 1574). A funnel graph is a scatter diagram of precision versus estimated effect (such as estimated elasticities, regression coefficients or partial correlation coefficients). Precision is best measured by the inverse of the standard error (1/Se). > As the name suggests, the expected shape is an inverted funnel — in the absence of publication selection. When there is no publication selection, estimates should vary randomly and symmetrically around the ‘true’ population effect. Because small-sample studies with typically less precision form the base of the graph, the plot will be more spread out there than at its top. However, it is the graph’s symmetry (or its absence) that is crucial for assessing publication selection (see Figure 1).
  - UnFleshedOne 7y ago
    Selecting correct presentation methods should be up to the platform, not the users.

pge 7y ago

The plot is showing that there is a negative selection bias in the reporting of stories about minimum wage. That is, that stories suggesting that increasing minimum wage has negative effects (increases unemployment, etc) are overreported relative to stories showing a positive effect from an increase in minimum wage. I am only repeating what the paper says; I don't claim to have validated their methodology or sources.

apcragg 7y ago
That is an incomplete summary. Their conclusion is that despite this larger than normal selection bias, the minimum wage appears to have no impact on unemployment. The paper's abstract, discussion, and summary all mention this as part of the central thesis - why leave it out?
- gowld 7y ago
  It's not an HN commenter's obligation to post everything a reader might want to see, certainly not highly technical point about the distinction between "this paper claims that evidence for an employment effect is over-reported" and "this paper claims that employment effect is approximately 0". They linked to the whole paper for the curious, and answered a particular question about it. Also, the use of "despite" does not logically fit the contents of (my) grandparent poster and parent poster, as grandparent poster didn't claim mention employment as a possible positive effect of minimum wage, and even within context of my parent post, "beyond" is a better word choice that n "despite"
- scarejunba 7y ago
  I think summaries (by their nature) must leave something out. Considering the abstract mentions what you said (and that it's reasonable for me to have read the abstract), I think it was reasonable for the responder to focus their attention on the plot alone (which was the focus of my question).
- IanCal 7y ago
  It is an incomplete summary of the paper, however it seems to be a pretty solid summary of what the plot is telling us.
- roenxi 7y ago
  The obvious starting position is that raising minimum wages has a negative effect on the number of low wage employees hired; because the money from higher wages has to come from somewhere and might not be available - leading to reduced business activity. This is also the classic economics position. With that in mind, there is going to be a bias towards publishing the 'surprising' results that there is no impact/a positive impact from minimum wages. The bias that this paper is identifying is surely a real thing, but it might potentially be a bais against, eg, poorly run studies that pick up spurious effects. So the conclusions the paper draws w.r.t. elasticity aren't that interesting; we'd need to know why the bias exists before it is particularly useful. I'm a bit suspicious of using 'meta-analysis' at all because that just means that the faction that repeats themselves the most with low-quality research gets to pick what the elasticity is. In theory we shouldn't need 200 studies to make a claim, we should need 1 or maybe a small suite (like, 5) of very high quality studies that intellectually honest economists struggle to rebut.
- pingyong 7y ago
  >why leave it out? Because he summarized the plot, not the paper.
scarejunba 7y ago
Much appreciated. And thank you to killjoywashere for selecting out the part of the paper that answers my question; gowld for reformatting appropriate to my mobile device. Apologies for not replying directly. I have a commenting rate-limiter applied to my account.
- Fjolsvith 7y ago
  I, too, have a commenting rate-limiter applied to my account. I can only post about 5 comments in a day before HN tells me to "Slow Down, You're Posting Too Fast".

killjoywashere 7y ago

From the paper: ‘The simplest and most commonly used method to detect publication selection is an informal examination of a funnel plot’ (Sutton et al. 2000a: 1574). A funnel graph is a scatter diagram of precision versus estimated effect (such as estimated elasticities, regression coefficients or partial correlation coefficients). Precision is best measured by the inverse of the standard error (1/Se). As the name suggests, the expected shape is an inverted funnel — in the absence of publication selection. When there is no publication selection, estimates should vary randomly and symmetrically around the ‘true’ population effect. Because small-sample studies with typically less precision form the base of the graph, the plot will be more spread out there than at its top. However, it is the graph’s symmetry (or its absence) that is crucial for assessing publication selection (see Figure 1).

michaelmrose 7y ago
This method of quoting is unreadable on mobile
- gowld 7y ago
  >‘The simplest and most commonly used method to detect publication selection is an informal examination of a funnel plot’ (Sutton et al. 2000a: 1574). A funnel graph is a scatter diagram of precision versus estimated effect (such as estimated elasticities, regression coefficients or partial correlation coefficients). Precision is best measured by the inverse of the standard error (1/Se). > As the name suggests, the expected shape is an inverted funnel — in the absence of publication selection. When there is no publication selection, estimates should vary randomly and symmetrically around the ‘true’ population effect. Because small-sample studies with typically less precision form the base of the graph, the plot will be more spread out there than at its top. However, it is the graph’s symmetry (or its absence) that is crucial for assessing publication selection (see Figure 1).
- UnFleshedOne 7y ago
  Selecting correct presentation methods should be up to the platform, not the users.