-1=e^ipi

You know you can always click on the image and it tells you were it's from, right? On the last page, I provided images from nature, skepticalscience and the ipcc.

For the sake of providing more information to people I'll briefly explain some of the basics of the Solow model and how it relates to optimal savings rates and consumption taxes. So a very basic idea of an economy is that the economy takes a combination of labour and physical capital to produce goods and services (i.e. if Y is economic output, L is labour and K is physical capital then Y = f(K,L), where f is some function). - Let's restrict f such that if you have more labour and more physical capital, then you have more goods and services. - Furthermore, let's restrict f such that if you keep either capital or labour constant and vary the other parameter then there will be decreasing marginal returns (basically, people benefit more from the first $100 of physical capital per capita than the next $100 of physical capital per capita). - Let's also suppose that there are constant returns to scale (i.e. if you double both K and L then Y will double). The simplest functional form that satisfies the above 3 properties is Y = A*K^a*L^(1-a), where a and A are constants. Furthermore, this functional form makes the prediction that physical capital's national share of income is roughly constant (this seems to be roughly justified by empirical evidence, see figure 3.3 from https://www2.bc.edu/~murphyro/EC375/PDF/Ch3.pdf).From the empirical evidence, physical capital's share of national income is approximately 1/3 and is equal to a. So in the very basic Solow model, you have some production function (we will use the one above) and the change in physical capital over a unit of time is going to be equal to the savings rate s, times the economic output Y, minus the depreciation rate d, times the physical capital K. Basically, new physical capital is created due to saving/investment and old physical capital is lost due to things wearing out (hammers, buildings, computers, etc. don't last forever; also the rate of physical capital being lost depends on the depreciation rate and how much physical capital you have). So in the long run equilibrium, the rate at which new physical capital is being created has to equal the rate at which old physical capital is being lost, so sY = dK. If we substitute in our production function, we get sA*K^a*L^(1-a) = dK. Now let k be physical capital per capita (i.e. K/L). If we divide both sides by L we get sA*k^a = dk. Isolating for k gives k = (sA/d)^(1/(1-a)). Let y be GDP per capita (i.e. Y/L), dividing the production function by L gives y = A*k^a. Substituting in the above value for k gives y = A*(sA/d)^(a/(1-a)). Now consumption per capita c is (1-s)*y (what you don't save you consume). So we get c = (1-s)*A*(sA/d)^(a/(1-a)). If you want to find the value of s that maximizes c, you have to ensure that the derivative of c with respect to s is zero. This implies 0 = -A*(sA/d)^(a/(1-a)) + (1-s)*A*(sA/d)^(a/(1-a))*(a/(1-a))/s. Dividing everything by A*(sA/d)^(a/(1-a))/s simplifies this to 0 = -s + (1-s)*a/(1-a). Isolating for s gives s = a. Thus to optimize long run consumption per capita, you want the savings rate to be equal to a, physical capital's share of national income, which is approximately 1/3. So let's look at some developed countries and their savings rate vs their consumption tax rate: Country: Savings Rate: Sales Tax: Norway 37% 25% S. Korea 35% 10% Taiwan 31% 5% Hong Kong 26% 0% Sweden 25% 25% Denmark 24% 25% Germany 24% 19% Canada 21% 5-15% USA 17% 0-7.5% UK 11% 20% South Korea and Taiwan are closest to 33% and have relatively low sales taxes (Hong Kong too), though I think that the main reason for the high savings rates in these countries is culture. Canada doesn't have a comparable savings rate even though it has comparable levels of sales tax. If we restrict ourselves to developed western countries then the 3 closest to the optimal are Norway, Sweden and Denmark, all of which have sales tax rates of 25%. This suggests that Canada should try to go for a ~25% sales tax (also 25% is nice because it is easily invertible). Edit: Also, the capital gains tax is very relevant here. Taiwan and Hong Kong have no capital gains tax. South Korea's capital gains tax is very low. Another thing to point out is that culturally homogeneous countries tend to have higher savings rates than culturally inhomogeneous countries. So this may be another reason why places like Norway and South Korea have a much higher savings rate than places like the USA or Canada.

The first one, but effectively both systems would be equivalent provided you adjust the guaranteed income and the tax rate accordingly. The first one is just a bit simpler.

Do you disagree with any of those 3 principles, Occam's Razor or empirical evidence?

KeepitSimple, a country can only have 1 tax system at one given time, so it makes sense to have the best possible tax system. The set of progressive taxes are infinitely dimensional. Even if you order all the possible progressive taxes by their 'progressiveness' and you wish to implement a tax with a certain level of progressiveness, that only gives you 1 criterion to choose the best tax system, which isn't enough to obtain a unique tax system from the set of progressive taxes. No, but I can think of a methodology that obtains a unique tax system (at least in terms of distribution of income tax) using the Pareto principle, the anonymity principle, the Pigou-Dalton principle, Occam's Razor and empirical evidence.

Btw, using the updated methodology, I get a median estimate of 14.64 GtC per year by 2100 under a no-mitigation scenario. For CH4 and N2O emissions, the divergence between RCP 8.5 is even more ridiculous. CH4 and N2O emissions per capita have been rapidly decreasing globally since the 70s, yet magically they are going to start increasing under the RCPs. With the same model I get median estimates of 167 Tg CH4 per year in 2100 and 5.35 Tg N per year in 2100. In RCP 8.5 we get:

I'll also point out that I did the two sided statistical test rather than the one sided statistical test (in which case the p-value would be cut in half). So I'm 96.32% certain that CMIP5 models are overpredicted temperature for 2006-2014, but only 92.64% certain that the observations are inconsistent with the CMIP5 predictions.

I explained my methodology here: http://www.mapleleafweb.com/forums/topic/24202-what-is-the-correct-value-of-climate-sensitivity/?p=1048951 And I made improvements to my methodology here: http://www.mapleleafweb.com/forums/topic/24202-what-is-the-correct-value-of-climate-sensitivity/?p=1049590

In order to get a p-value of 0.5, the z-statistic needs to reach 1.96 or -1.96. So if the trend continues, the the p-value will reach 0.05 in approximately (1.96/(1.7894/sqrt(9)))^2 = 10.8 years. So CMIP5 predictions might be falsified in 2 years time.

Please define what you mean by 'beat upon' and explain how I 'beat upon' Cowtan and Way. Source? The source is the calculations I did in this thread. I explained the methodology, data sets, etc. See my above posts. RCP 8.5 is the most representative from 2006-2014. That doesn't mean RCP 8.5 is representative of BAU; RCP 8.5 makes assumption after assumption in favour of extreme warming. It isn't a representative BAU scenario, it is an alarmist scenario. Fitting an exponential trend to CO2 emissions per capita and a logistic trend to population demonstrates strong divergence between RCP 8.5 and what you would expect under BAU. The temperature projections are based upon CMIP5, and I was testing the hypothesis of if CMIP5 predictions are validated by the instrumental data. I already explained a number of reasons why I think that CMIP5 models are oversensitive and thus are overestimating future warming.

Not always. But if the person would get more utility from having the monetary value of an hour's work then they would from having an extra hour of leisure time, then generally it would be preferable if they worked that additional hour. I'm not. I think we should have the best tax system possible. And it's unfair if the two fishermen don't pay the same amount in taxes.

Tiered flat rate isn't a flat tax. And what prevents you from doing this?

Okay, using GetData on this: I get that the warming from 2005-2014 under RCP 8.5 is approximately 0.00558 C larger than under the mean of the RCPs. If I assume that the difference in median temperature prediction between RCP 8.5 and the mean of the RCPs is roughly proportional to time for 2005-2014, then I should adjust my displacement values to take this into account. I get (I'll assume that the standard error does not change by restricting things to RCP 8.5; even though this restriction should reduce the standard error; I'll give the benefit of the doubt to the claim that model predictions agree with observations): Year: Displacement: SE for prediction: SE for observation: Total SE: 2005 -0.0089 0.1364 0.0300 0.1397 2006 -0.0702 0.1157 0.0290 0.1193 2007 -0.0699 0.1190 0.0290 0.1225 2008 -0.2229 0.1151 0.0300 0.1189 2009 -0.1058 0.1155 0.0300 0.1193 2010 -0.0536 0.1207 0.0310 0.1246 2011 -0.2217 0.1198 0.0310 0.1237 2012 -0.2286 0.1234 0.0310 0.1272 2013 -0.2396 0.1285 0.0310 0.1322 2014 -0.1923 0.1346 0.0310 0.1381 Year: Displacement - 0.5846*Previous_Displacement: SE for first displacement: SE due to second observation: SE due to autocorrelation constant: Total SE: 2006 -0.0650 0.1193 0.0229 0.0062 0.1216 2007 -0.0289 0.1225 0.0222 0.0175 0.1257 2008 -0.1820 0.1189 0.0222 0.0175 0.1223 2009 0.0245 0.1193 0.0229 0.0313 0.1255 2010 0.0082 0.1246 0.0229 0.0215 0.1285 2011 -0.1903 0.1237 0.0237 0.0153 0.1269 2012 -0.0991 0.1272 0.0237 0.0312 0.1331 2013 -0.1059 0.1322 0.0237 0.0317 0.1380 2014 -0.0523 0.1381 0.0237 0.0324 0.1438 Normalized: 2006 -0.53416 2007 -0.23005 2008 -1.48847 2009 0.19530 2010 0.06372 2011 -1.49935 2012 -0.74405 2013 -0.76786 2014 -0.36327 The mean is now -0.59646 and the z-statistic becomes -1.7894. The p-value is now 0.0736. Lower, but not enough to make the difference statistically significant. I can't reject the hypothesis that model predictions agree with empirical data.

I'll also point out that Emissions over 2006-2014 were higher than all of the RCPs except 8.5: So only RCP 8.5 can be representative of what happened over the past 9 years, where as if CMIP5 was making non-biased predictions then the results from running the other 3 RCPs is expected to be lower than observations. Figure 11.25 includes all RCPs, thus the confidence interval is larger and lower than it should be based on emission rates from 2005-2014. If you account for the fact that only RCP 8.5 accurately reflects what happened since 2006, then that p-value of 7.97% will decrease and should easily go below the 5% significance level. Thus CMIP5 predictions probably do not agree with empirical observations (I need the RCP 8.5 confidence interval to properly test).

I stand corrected with respect to CMIP3 projections using the A1B scenario in 2000 being falsified by temperature trends. For the second post, even with the update to HadCRUT 4.3, you can't say that observations aren't outside the 95% confidence interval. For example, the median temperature estimate for 2011 is clearly below the 95% confidence interval for 2011. Anyway, while seeing if observations are inside the 95% confidence interval might be a simplistic way to check by eyeball if the data as validated the model predictions or not, it isn't the correct way to test the hypothesis. It is possible for the data to validate the model even if an observation lies outside the 95% confidence interval (for example, if you have 20 observations of annual temperature, then chances are at least 1 year will be outside of the 95% confidence interval) and it is also possible for the data to falsify the model even if no observations lie outside the 95% confidence interval (for example, if the majority lie at the bottom of the confidence interval). So let's properly test the hypothesis of if the RCP projections are validated by observations. Now the 95% confidence intervals above are all annual confidence intervals. It would be nice to assume that observations in each year are independent, but that would be unfair since residuals around the temperature trend clearly have an autocorrelation. So I will need to first determine the autocorrelation of temperature residuals around their trend. So let's take 1850-2005 for Cowtan and Way version 2.0: http://www-users.york.ac.uk/~kdc3/papers/coverage2013/had4_krig_annual_v2_0_0.txt. First I'll obtain of the residuals after removing a quadratic trend. Next I'll fit the AR(1) autoregressive model without constant to the residuals. I get an autocorrelation coefficient of 0.5846. The next thing I need are the bounds of that 95% interval for each year. Unfortunately, I can't find it in a text or excel file or something, so I'll obtain it from that image you provided using a program known as GetData + Linear interpolation. I get: Year: Lower Bound: Upper Bound: 2005 0.0233 0.5581 2006 0.0756 0.5291 2007 0.0953 0.5617 2008 0.1183 0.5693 2009 0.1298 0.5824 2010 0.1398 0.6129 2011 0.1710 0.6405 2012 0.1922 0.6760 2013 0.2086 0.7123 2014 0.2247 0.7524 Now, let's turn this into the median estimate and the standard error for sake of later calculations (assuming a roughly normal distribution here; the standard error corresponds to the length of the confidence interval divided by 3.92): Year: Median: Standard Error: 2005 0.2907 0.1364 2006 0.3024 0.1157 2007 0.3285 0.1190 2008 0.3438 0.1151 2009 0.3561 0.1155 2010 0.3764 0.1207 2011 0.4058 0.1198 2012 0.4341 0.1234 2013 0.4605 0.1285 2014 0.4886 0.1346 Now the Cowtan and Way data needs to be converted to anomaly from 1986-2005 baseline to be comparable. This gives: Year: Temperature Anomaly: 2005 0.2818 2006 0.2328 2007 0.2598 2008 0.1228 2009 0.2528 2010 0.3258 2011 0.1878 2012 0.2098 2013 0.2258 2014 0.3018 Now let's obtain the displacement of observation from the median by year. I'll also list the standard error after you include the observational uncertainty in global temperature according to Cowtan and Way: Year: Displacement: SE for prediction: SE for observation: Total SE: 2005 -0.0089 0.1364 0.0300 0.1397 2006 -0.0695 0.1157 0.0290 0.1193 2007 -0.0687 0.1190 0.0290 0.1225 2008 -0.2210 0.1151 0.0300 0.1189 2009 -0.1033 0.1155 0.0300 0.1193 2010 -0.0506 0.1207 0.0310 0.1246 2011 -0.2180 0.1198 0.0310 0.1237 2012 -0.2243 0.1234 0.0310 0.1272 2013 -0.2347 0.1285 0.0310 0.1322 2014 -0.1868 0.1346 0.0310 0.1381 Now I can't just assume each displacement is independently distributed. There is clearly auto correlation. I.e. the displacement for 1 year should equal ~0.5846 times the displacement of the previous year plus a residual. So I want to look at the displacement for 1 year minus 0.5846 times the displacement of the previous year to see if the observations agree with model predictions or not. Now to be fair there is uncertainty associated with the autocorrelation constant 0.5846 (it has a standard error of 0.0662) and I also need to include the uncertainty of the observation of the previous year: Year: Displacement - 0.5846*Previous_Displacement: SE for first displacement: SE due to second observation: SE due to autocorrelation constant: Total SE: 2006 -0.0643 0.1193 0.0229 0.0062 0.1216 2007 -0.0280 0.1225 0.0222 0.0175 0.1257 2008 -0.1808 0.1189 0.0222 0.0174 0.1222 2009 0.0259 0.1193 0.0229 0.0311 0.1254 2010 0.0098 0.1246 0.0229 0.0213 0.1285 2011 -0.1884 0.1237 0.0237 0.0149 0.1269 2012 -0.0969 0.1272 0.0237 0.0309 0.1331 2013 -0.1035 0.1322 0.0237 0.0314 0.1379 2014 -0.0496 0.1381 0.0237 0.0321 0.1438 Okay, now let's normalize everything to Displacement - 0.5846*Previous_Displacement divided by its standard error: 2006 -0.52906 2007 -0.22309 2008 -1.47946 2009 0.20645 2010 0.07658 2011 -1.48496 2012 -0.72813 2013 -0.75069 2014 -0.34482 Now if the the predictions agree with observations, then the mean of the above values should be zero. The mean is -0.58413. We can test if predictions agree with observations using a simple Z-test. Given that there are 9 observations, under the hypothesis that predictions agree with observations, the probability distribution of the mean should have a mean of zero and a standard deviation of 1/sqrt(9). The z-statistic is thus -1.7524. The p-value of this z-statistic is 0.0797. So I cannot reject the hypothesis that predictions agree with observations at the 5% significance level, though it's pretty close.

The misrepresentation continues. biasedness != reliability. small != insignificant. Keep pretending otherwise if you wish.

Not on a percapita basis. https://www.fcc-fac.ca/en/about-fcc/media-newsroom/news-releases/2014/fcc-report-shows-canada-is-worlds-top-per-capita-ag-trader.html

We actually have quite a few advantages based on our geography, population density, technology, etc. We can be agricultural exporters without subsidies. Also, given that this discussion started with the mention of supply management, I'll point out that supply management only affects a few agricultural industries such as eggs, milk and poultry. Maybe less profits in the agricultural sector, but lower taxes elsewhere, which means more profits elsewhere to offset it. Unless there are positive externalities associated with having an agricultural sector, the costs exceed the benefits.

I mean you would have to get rid of all taxes, not just on poor people, for the property of a single tax rate to be satisfied. Unless you meant something different by single tax rate than what I would take that to mean. Why can't you use the lump sum transfer to pay rent or buy food? I don't get it. I'm not really sure if 'cause other countries do it' is a good justification. If all other countries implement a bad policy, it will still be bad even if Canada implements it. If you implement subsidies, you have to tax other industries to pay for it. So you have to see if the overall result is beneficial to Canada or not. Without some sort of justification such as externalities or common/public goods I don't see how implementing agricultural subsidies makes sense.

And people would still have an incentive to work provided the tax rate is less than 100%. Me and my ilk? Who are these ilk exactly? Please find me one example where I was complaining that 'handouts make people lazy and refuse to work'. Maybe, but are direct subsidies justified? What is the reason for market intervention in the case of the agricultural sector? Externalities? Lack of information in the market? Public goods? Common goods? I guess you could argue that there are positive externalities associated with nutrition and having a country have an ability to feed it's people has national security value (public good), but still the justification seems pretty flimsy.

Actually, I think it's closer to $11,000 per year for me. And I eat a lot of eggs (cheap source of protein). It would be nice if we got rid of supply management though.

Yes.

I wasn't trying to emphasize the timing, but given that some people are bad at saving money, it would probably make sense to just give everyone weekly direct deposits from the government.

1 and 3 are in direct conflict, unless you advocate no taxes. Having a flat tax with a lump sum transfer satisfies 3, while effectively satisfying 1. They can afford it because the tax rate would be somewhere between 0% and 100%. And if you make it revenue neutral, the lump sum transfer will be larger than what poor people pay in taxes.

Can't they just donate to charity? Anyway, if people want to pay effectively more tax than they have to, I have no opposition to that. It would probably still make sense to have a consumption tax and pigouvian taxes. Depends where you live I guess. My rent + utilities is $500 per month. Food doesn't have to be expensive if you know where to shop. For transportation, you can always try to carpool, walk/bike or use public transportation. I have internet access. This counts for all 3.