Friday, September 28, 2018

Sexual Assault in the 80s and Christine Blasey Ford's Testimony


Dr. Ford’s testimony alleging that Judge Kavanaugh attempted to rape her when she was 15 was extremely difficult to watch as she was and still is deeply traumatized by the event. Kavanaugh’s abrasive and highly rehearsed obstructionist response was even more difficult to watch. In response I decided to look at public data to see if the type of attack Ford described was typical of that observed by social scientists.


I used the Longitudinal Study of Violence Against Women (Citation below). This study does not match exactly her cohort but it is close as it follows women who were born between 1972 and 1973 while Ford as born in 1967. But one should think the threats of violence against women at that time were similar to those faced by Ford. The study interviews women first entering public colleges at ages 18 during their orientation then asks them follow up questions at ages 19, 20, 21, and 22.


Over each subsequent interview period more and more of the sample dropped out of the study leaving only 47.9% who completed the full study. That said, the study is detailed with hundreds of difficult and intrusive questions that tolerant students responded to.


Ford’s Testimony
First off I would like to look at childhood sexual assault data. Was Ford’s experience typical of those suffering this kind of trauma or no?


She reported to be aged 15 at the time which is less common as the risk of such trauma increases with age.



She reported that Kavanaugh attempted to force her to have sex with him. Most of the respondents to the study fortunately reported that they had not had that experience.



Ford reported that Kavanaugh was somewhere between a casual acquaintance and a friend. She had been at several events with him in the past. One of the boys that she had gone out with had introduced him to her.


Finally, during the dreadful event, how did she communicate her unwillingness? Ford use both physical resistance as well as screaming to resist the attack. Hearing her describe her  sense of fear helplessness as he placed his hand over her mouth was devastating.


A general study such as that which I am referencing can have little if any bearing evaluating the truth of an individual claim. However, it is worth noting that of the women who responded to this study and the millions they are sampled from, should and would be able to personally to relate to the terrible events that happened these 37 years ago.


As a man largely personally insulated from such violent and devastating aggression, I found myself deeply moved by Ford's testimony. Reviewing this data personally has been difficult for me.


General Trends


Use of force to Coerce Sexual Activity


Looking more generally at the data in the longitudinal study you can see women who are first interviewed at 18 slowly or quickly move from being yet unvictimized to becoming victims through their college years.

The above figure shows four variables measured at time of entry into college and each year after. The initial year for each variable does not list a number but each year after is numbered sequentially. The chart shows self reported rates from top to bottom of actions women were forced to perform or nearly so through the use of force by males: 1 Sexual acts such a oral or annual sex, 2 Sex play, 3 Intercourse, and 4 unsuccessful attempts at intercourse.


The need for imputation was due to a high drop out rate from the study especially among those who reported such acts. In order to generate the imputed numbers I just assumed that anybody that reported any of the above activities in previous years would report them in future years if they had been interviewed. As a result all of these numbers are under-representational of the actual rates of sexual assault. To calculate the percentages that these values are scaled from, I divided each total number of those being reported by the total sample size 1580.



Use of other coercive means

I also looked at the use of other coercive means in order to obtain or attempt to obtain sex from an unwilling partner.
The interviewers asked the women if they had engaged in sex because a figure of authority forced them to through their power or position (variable 1) or if someone in authority had attempted to (variable 2). Women seemed most likely to engage in intercourse or sex play because of excessive unwanted attention from men (variables 3 and 4). Men also were successful or attempted to obtain sex through use of excessive alcohol or drugs (variables 5 and 6).


Conclusion
The environment in which Dr. Ford grew up in was a dangerous environment for young women in which men often preyed upon the vulnerable and naive. Obviously only Ford, Kavanaugh, and Mark Judge know what happened in that room. Ford has undergone a lie detector test and requested professional investigators look into her claims while Kavanaughand and Mark Judge have resisted any kind of criminal investigation and refused to submit themselves to a similar polygraph test.

Code on Github

Data

White, Jacquelyn W., University of North Carolina-Greensboro, and Humphrey, John A. Longitudinal Study of Violence Against Women: Victimization and Perpetration Among College Students in a State-Supported University in the United States, 1990-1995. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2015-09-11. https://doi.org/10.3886/ICPSR03212.v1

Wednesday, September 26, 2018

Strategizing Retirement Investments (In the US)

Major caveat! I have no investment training or finance training and all of my calculations are back of the envelope calculations put together from what information I can gather online. In addition, given that investment planning usually spans decades, massive uncertainties exist in tax schemes and expected rate of returns. Please consult a professional!


That said, as someone who has an income you are trying to decide if and how much to invest in your retirement account. Most people have several options: tax deferred retirement options such as traditional IRA, 401(k), 457, or 403(b). Some employers also offer matching programs which match contributions at a rate of 50%, 100%, or more to a limit each year. As tax deferred retirement accounts, money contributed to them are not taxed at the time of contribution (except for Social Security tax 6.2 and Medicare 2.9).

In addition to these programs there is the option of investing in a Roth IRA with income that has already been taxed but will suffer no additional capital gains tax. Finally, there is the option of investing directly (with pretaxed earnings) such as purchasing stocks, mutual funds etc. While direct investments give you the freedom to withdraw money at any time, they suffer the penalty of being liable for capital gains taxes (15% on nominal gains for long term investments).

Each of these investment choices has its relative merits. One thing to consider is tax bracket.  Each earning level of taxes are paid at a specific bracket. For example, for a couple the first 19k of earnings is paid at a 10% rate, while the next 58k of earnings is paid at 12%, while the next 88k is taxed at 22%, and so on with increases to 24%, 32%, 35%, and finally 37% at total earnings above 600k (see tax tables). Yet the combined effect of these changing marginal tax rates is to turn the effective tax rate into a globally non-linear function (Figure 1).

For the individual the federal tax rate for a median household income level $32K or for a married couple making $59k is about 11.5%. The overall effect of tax bracket structure is to overstate the tax obligation that most individuals and couples are required to pay. For example, a household making 103k (75 percentile) while paying 22% on their last dollar earned are overall only paying only around 14% of their total income in income tax. Even households making 250k (97 percentile) while their tax bracket is officially 24% on their last dollar earned are only paying a tax rate a little above 19% on overall income.

Figure 1: Overall Federal Income Tax Rate without Social Security and Medicare
That said, nobody likes to see their income disappear to taxes so I am not trying to say taxes are too high or too low. I am just noting that tax brackets alone for the vast majority of the population, are quite inaccurate tools at estimating tax liability. If we include the social security tax and Medicare (Figure 2), federal taxes become both more painful overall and less progressive (higher earners paying higher taxes). This figure is even more non-linear with important portions of the figure (earnings above 128k) experiencing a lower tax rate growth than earnings up to that point. Curiously, an individual income earner above 128k actually experiences a decline in overall tax rate until earning up to the next tax bracket of 157.5k.
Figure 2: Overall Federal Income Tax With Medicare and Social Security

But enough of this! About 75% of households are earning 110k or less. Those earning above that rate are likely already talking with financial managers in order to figure out how to plan their futures.

The overall point of the above discussion is that tax differed savings plans might be quite beneficial for retirees as many retirees expect to be at a lower income level than they were while working. Income drawn from such plans is taxed at their current income level rather than the level they had at the time of saving. Social Security and Medicare tax is always paid immediately upon all wage earnings (and thus constant under all scenarios) so it will not be included in the following analysis.

In order to analyze the benefit of choosing different investment mechanisms, I looked at three different factors, 1-overall effective lifetime income, 2-total investment worth at the end of retirement (inheritance available), and 3-annual effective income. The first two seem like they are pretty important for most people but arguably annual effective income is probably the most important factor. Ideally, retirees looking to enjoy their retirement will not suffer a minimal loss in purchasing power after retirement.

In order to look at different investment strategies, I simulated two earning levels, 55k per year gross and 110k per year gross. Workers work for 35 years with real annual salary increases of 0.5%. After 35 years of work, workers consider the next 20 years as years living on secondary income sources (1/3 of base income) as well as retirement savings withdrawn at a rate of 4% per year.

In order to make the numbers make sense, I attempted to simulate “real” money rather than nominal (including inflation). As a result I assume that tax brackets remain the same in terms of “real” earnings. Annual raises are modest as a result as well. In my simulation, I include a real investment rate of return of 5% and a capital gains tax of 20% on all direct investment withdraws. This does not take into account in the cost of such investments (which is tax free) while including a higher rate of capital gains tax to take into account the false effect inflation has on direct investments (and the corresponding tax on those investments).

In addition of the tax differed retirement accounts (traditional IRA or a 401k without matching – 401km00), the after tax but not capital gains taxed retirement accounts (Roth IRA), and the direct investment mechanisms (Direct), I also included two investment options  in which employers match contributions by either 50% or 100% (401km05 and 401km10).

Generally, all of the investments are considered on the basis of “investing” a certain amount of money 5.5k or 11k. There was one scenario in which I allowed for a greater investment quantity in order to allow the simulation of investing the same amount of after tax income (i401k00b). When the earning level was 55k and the investment was 5.5k for i401k00b the 401(k) investment level was instead 6.5k. And when the earning level was 110k and the investment was 11k for i401k00b the 401(k) investment level was instead 14k.

In my analysis I am concerned with "effective income" which I define as:  
effective income = all income – tax – investment contributions (only made for years working).

Gross Base Income 55k per Year
Under the different investment scenarios, total lifetime earnings vary between 2.3m and 2.8m in the case when employers provide matching funds (Table 3). When no matching funds are provided, lifetime earnings are more similar between scenarios with lifetime earnings between 2.3m and 2.5m. Differences are more exaggerated when considering total assets at the end of the retirement period included in total lifetime earnings with the highest level obtained under scenario i401km00b (no matching but investments made at a rate that are higher than 5.5k, but instead meant to achieve after tax 5.5k investment per year).
Figure 3: Total Lifetime Effective Earning and End of Retirement Assets for Gross Income of 55k per Year
Rather than look at lifetime effective income and total assets it might be better to look at per year income (Table 4). Without contribution matching, all of the scenarios result in a significant drop in effective income. The most dramatic drop is in the "No Saving" scenario in which the effective income of the worker drops from near the global maximum for the scenario to the global minimum of the scenario as the worker enters retirement without any additional income sources outside of the 1/3 base income (SS income or whatever) provided under all retirement scenarios.

Looking at the effective income while working levels, there are three levels that the simulated worker experiences. The highest being "No Savings", which results in the highest possible current effective income level. The second are the pretax saving schemes such as 401km10, 401km05, and 401km00. These schemes provide highest effective income during working years because they reduce the after tax income not by 5.5k but 5.5k after taxes which is 4.8k in year 1. The scenario in which the highest matching funds are available is that which produces the highest long run returns. The third effective income level on working years is the direct investment scenario, the Roth IRA scenario, and the 401km00b scenario (which is custom calibrated to achieve this). In this income scenario 401(k) investment even though it is taxed upon dismemberment produced higher earnings than that of the Roth in retirement though the difference does not appear to be large in this scenario.
Figure 4: Annual effective income under different investment mechanisms 55k gross income per year. Note a jitter is added to lines to aid in deciphering overlap.


Gross Base Income 110k per Year with 11k annual investment
Figure 5: Total Lifetime Effective Earning and End of Retirement Assets for Gross Income of 55k per Year

Overall, doubling the household income and the investment level results very similar relative distributions of Total Lifetime Effective Earnings or Total Assets at retirement.
Figure 6: Annual effective income under different investment mechanisms 55k gross income per year. Note a jitter is added to lines to aid in deciphering overlap.

Likewise when looking at lifetime smoothed effective income, the overall returns from different investment mechanisms remain similar. That said the difference between the investment strategy using the Roth IRA and increased investment using 401km00b (401 tax differed investment without any match at 11k after tax investment level which is 14k before tax) resulted in the same level of effective income during working years but significantly higher effective income in retirement.

That said, Roth IRA funds are more flexible than those of tax differed funds which suffer a 10% additional penalty if withdrawn before age 59.5. In addition, if you believe that your current income level is likely to be lower than your future income level, it probably means that you should consider investing more in your Roth over that of a tax differed investment instruments.

Comparing 55k to 110k per Year
The 110k per year scenario is basically a doubling of the 55k per year scenario in terms of both income and investment (except in a slight difference with the 401km00b scenario). Comparing the average annual effective income under these two scenarios we can infer the cost of higher income taxes in the 110k scenario. The average ratio is 1.94 which indicates a difference of .06 or 6%. Looking closely at Figures 1 or 2 we can see that the difference in overall tax rates between households making 55k per year and 110k per looks to be about 6% as well. Using larger tax differed investments (401km00b) does appear to be a way to achieve a smaller tax penalty with a decreasing in effective tax rate of closer to 4%.

Table 1: Shows the ratio of average annual effective income (same as total lifetime income). If the ratio were 2 then there would be no tax different between earners at 55k per year and 110k per year.

Ratio
NoSavings 1.94
401km00 1.95
401km00b 1.96
401km05 1.93
401km10 1.92
Roth 1.94
Direct 1.93
Average 1.94

Conclusion
If there is a 401k or other tax differed investment option in which matching is available, I would suggest putting the maximum your finances allow. If matching funds are not available, in this simulation investing pretax income in the 401k equal to that you would invest in the Roth IRA appears to yield a slightly better outcome. This is generally because while these funds will be taxed when withdrawn, having that much more funds grows that much faster than the after tax funds otherwise available.

Finally, in almost all of the scenarios, there is a significant loss of effective income upon retirement. If you would like to mitigate the risk of that happening, combining investment mechanisms is likely to be the best strategy. This might mean relying upon direct investments if for instance you have already filled your Traditional IRA or Roth IRA and have no other investment mechanisms provided by your employer.

Feel free to explore the simulation on your own with your own earnings and investment options! Code that produced this analysis can be found on Github

Wednesday, September 12, 2018

Plotting the Impact of Atlantic Hurricanes on the US

With hurricane Florence bearing down with expected to be devastating force, it might be a good time to reflect on the history of Atlantic Hurricanes on the United States. As an easy if not full proof source for hurricane data I drew on public data collated through Wikipedia's list of costliest Atlantic hurricanes as well as linked articles. On the page it lists 45 hurricanes going back to Hurricane Betsy in 1965, with 30 of them since 2000 totaling 803 billion dollars in estimated damages and 6,591 US fatalities. While this is quite a large number of fatalities, the death toll of Atlantic Hurricanes outside of the US is significantly larger with nearly 30,000 deaths estimated.

Within the US, the track record for deaths from hurricanes tend to be quite low with all hurricanes recorded having a fatality rate of less than two hundred with the exception of Maria (2017 - Trump administration) and Katrina (2005 - GW Bush administration).

Figure 1: US Fatalities by Year.

I have also scaled the y-axis by log10 to make it easier to examine less deadly storms.


Figure 2: US Fatalities by Year Log10 scaled.


Maria and Katrina being the most deadly of storms does not indicate that they are necessarily the most damaging of storms. 


Figure 3: Damages in the Billions by Year.
Harvey and Katrina are reported to have done 125 billions dollars each followed by Maria doing damage of 91.2 billion, Sandy at 68.7, and Irma at 64.8.

So why is it that Harvey (2017 - Trump), a storm that did as much damage as Katrina resulted in only 106 deaths while Maria a storm that did 25% less damage resulted in a US citizen total death toll comparable to the sum of all 44 other Hurricanes on record (2982 vs. 3609)?


Figure 4: Total deaths by Year.

We might attempt to predict which storms might be the most deadly using the storm measurements highest wind speed recorded and lowest pressure recorded.

Figure 5: Wind speed by US_Fatalities
Figure 6: Lowest air pressure recorded in terms of mbar (whatever that is).

As we can see that Maria and Katrina are both on the higher end of  the wind speed distribution and the lower end of the air pressure distribution, we can assert that a more powerful storm is a necessary component for the kind of death tolls associated with these storms.

What factors must be present for a severe storm which otherwise would be associated with a death toll of less than 200 people to being a tremendous human tragedy, such as in the case of Katrina or even more so Maria are yet unknown given the sparsity of data available to the public.

From the data it appears that Atlantic hurricanes have grown more damaging and more deadly for US citizens.

Hopefully, despite the catastrophic blunders of the handling of Katrina and Maria by the Bush and Trump administrations respectively, future hurricanes such as Florence will be dealt with in a more responsible manner.

Code and data can be found here