I voted online so you don’t have to


Alaskans can legally smoke marijuana, never pay state taxes, and vote online. Because freedom is something we take seriously and I know how to use computers, this year I chose to digitally vote as part of what the Council of State Governments calls an online voting laboratory. This is real voting (President, Senate, etc.) on real computers.

Alaska is the only state that allows any voter to cast their ballot online. Several other states allow some form of email or electronic voting, according to the National Conference of State Legislatures, but those jurisdictions mostly restrict e-voting to members of the military who are stationed overseas and are covered by the federal Uniformed and Overseas Citizens Absentee Voting Act (UOCAVA), and a few select others. Since 2012, Alaska allows anyone to vote online. I wasn’t an early adopter of smartphones or FitBits, but I was determined to be among the first in our country to vote from the comfort of my bed on my dumpster-salvaged laptop.

Early October: “Let me Google this.”

I start with Google, which tells me that no states whatsoever allow online voting.
But I’ve been on the Division of Elections website enough in the past several years to know that there is an option vote to by fax and some form of electronic voting. The DoE page soon (in no fewer than 915 words) tells me that I can apply for e-voting two weeks before the election. But no sooner.

Monday, October 24: Go Time.

Today is the first day possible to apply for an electronic delivery ballot or even access the application. The instructions are available on a video beforehand, but if you want a head start, you’ll have to wait until today. But the “online” dimension of this soon begins to show its cracks. Early Monday morning I began my online voting by downloading and printing a two-page PDF.

You fill out a few questions, sign with a pen (digital signatures will be rejected), scan, and then email or fax the complete application to the Division of Elections. As I soon learn, you essentially have to work in an office setting or have a stellar all-in-one scanner-printer-fax machine at your house to be able to vote online. There is a lot of paper involved. At 1:31 p.m., I email my application to the DoE.

Tuesday, October 25: The suspense is killing me.

With the exception of an auto response Monday from the Division of Elections system, there is no word on my application. I wanted to get this thing done Monday, but I instead comfort myself by reading 4,904 words about how the 538 election model weights different state polls and factors a suite of economic fundamentals into their projections. You should too.

Wednesday, October 26: I’m in.

At 6:08 a.m., I receive a 661-word email from DoE letting me know that I can vote within 12-24 hours. Or at least that’s what I understand, based on the somewhat muddily-worded message:

“Email Notification of ballot availability through the Online Delivery System will be sent to you in the next 12 to 24 hours”

At 8:11 a.m., I open another email letting me know that I can access the “voter portal” at https://ak.secureballotusa.com. Game on, time to vote this election and be done with it. But this message is not from the Division of Elections, it’s from a different email connected to a non-governmental domain: alaskaelections@secureballotusa.com.

The same State of Alaska that depends on a 30-year-old mainframe computer didn’t go out and build a secure stack of software to manage a relatively groundbreaking internet implementation of our democracy’s most important function.

They hired a contractor.

Cómo Votar: I enter the rabbit hole.

Secureballotusa.com redirects to the Barcelona, Spain-based company that implements e-voting systems across the globe. Indeed, the domain name is linked to a Barcelona address for Scytl Secure Electronic Voting, the contractor behind Alaska’s system.

Their Spanish headquarters are surrounded by small motorcycles.

With a tagline of “Innovating Democracy,” Scytl is the global frontrunner in electronic voting systems–their CEO claims a 90% market share–and the contractor hired in 2012 by Alaska to build a secure system for electronic elections. Formed by cryptographers in 2001, the company has experienced massive growth and plans to have 800 employees by the end of the year. The firm clearly sees a global market: Scytl has teams based in Dubai, Hong Kong, London, Ukraine, and Bogota, among other locations. The American branch is registered in Virginia and has offices in Tampa, Florida and Oklahoma City. And they’re expanding fast, according to their website.

“Over the past years, Scytl has built a strong financial reputation thanks to a substantial growth rate ranging from 65% to 70% per year and the support of leading financial investors: Balderton Capital, Nauta Capital, Spinnaker Invest, Vulcan Capital, Sapphire Ventures, Vy Capital, Industry Ventures and Adams Street Partners.”

Who owns this thing?


An early tech billionaire and an international team of venture capitalists from Silicon Valley to Spain to Dubai have stakes in the company and are betting with their funders that the global trajectory of software eating the world will continue with state, local, national, and corporate elections. The company already does business with the UK (Brexit poll worker training), the Swiss Canton of Neuchâtel, the Norwegian Government, some sort of businessman-backed eDemocracy network in Russia, as well as Spanish McDonalds restaurants.

Not far from Sand Hill road, the Sapphire Ventures team leadership captures the flavor of international money behind Scytl. According to their website, Nino Marakovic “was born in Croatia, is an Austrian citizen and has also lived in Italy and Germany before settling in the San Francisco Bay Area more than 15 years ago.”


Across the Atlantic, Romanian-born former Goldman Sachs Banker Alexander Tamas, a past partner with the Russian Digital Sky Technologies now runs a Dubai-based venture capital and private equity firm that is among the investors in Scytl. They’re joined by Spinnaker SCR – a Spanish private equity fund under the Riva y Garcia group that invests in media companies that focus on “…the production, distribution and presentation of all kinds of contents, as well as the generation of application technologies for the sector…”

Closer to the Pacific Northwest is Vulcan Capital, an investment group led by Microsoft co-founder billionaire Paul Allen. His team is active in e-commerce in the Middle East and have within their portfolio a virtual reality play, a space launch startup, a small-molecule genetics biomedical firm and of course, Uber. They are joined by Industry Ventures, another Silicon Valley venture capital firm with bets on Twitter, Uber, and dozens of startups with bad logos.


The internet quickly turns George Soros-weird upon looking more into Scytl involvment in the United States. The web is full of allegations that Obama gave the company a processing contract for the US voting system (there is no federal voting system), and that George Soros is an owner. The company lists its investors online and Snopes found no evidence of Soros as an owner. The company is planning a 2017 IPO on the NASDAQ stock exchange, reports NovoBrief, shooting for a one billion dollar valuation, according to ASCRI.

The company points out that they have systems for avoiding conflicts of interest in their team. “In compliance with Scytl´s strict political neutrality policy, none of Scytl’s shareholders or senior management members have any political affiliations,” the website states.

As a citizen, I’d suggest it’s worth having at least a passing familiarity with the billionaires associated with your public elections.

Thursday, October 27: But first, some Election humor.

I open the link in my email, supply some relatively basic identifying data (last four digits of Social Security Number) and enter the world of election software jokes in the training materials.


For those keeping track, if the ptarmigan designation were to pass, this would add ambiguity to the status of our current state bird, the willow ptarmigan.


The video is currently enjoying the stellar review record on YouTube of four thumbs up and ZERO thumbs down.

There is one moment that makes use of a clip-art calendar in a rather aggressive way, showing a conceptual election day on the fourth Tuesday of the month (remember to vote November 22 everyone!)


Another strange moment happens when the woman’s doppelganger appears to witness her signing her voter certificate.


There is a shocking amount of origami involved if one is to return the electronic ballot by mail (which is indeed an option).


And finally, whoever runs the company’s YouTube channel also must have gone on an Easter candy-fueled video liking spree:


Upon reaching the actual voting page, it’s a bit surreal: I’m actually going to vote online. On the computer. From my bed.

The language gets a little muddy; I don’t actually “download” my ballot here, I enter the online interface to fill out the ballot.


You simply click on the oval. This is voting. No butterfly ballot and no autocorrect for this portion of the app.


You then cycle through your statewide races, local Alaska Legislature House and Senate races, Ballot Measure One and then on the retention of something like 26 Alaska judges, which you have obsessively researched over the nine months and are fully qualified to vote on. Once you’re done, you download the two files (PDF #2 and #3 of the process for those keeping track).


Friday, October 28: Let’s ship this thing.

I log back in to upload the ballot and certificate. I figured the system may give me some trouble because I left after completing my ballot, but before uploading the final document. No such trouble.

Here’s where computer literacy comes into play. If you’ve gotten this far, you’ll probably make it home free. But this isn’t one stop shopping online: you have to download, locate, unzip, print, sign, scan, and upload more documents before you’re done. If you run into trouble, the system provides a phone number with a South Dakota area code.


This is in no way faster or easier than voting in person. It’s like doing your taxes on paper for the challenge of it. For those who are used to being able to manage their financial and social lives completely online, this is not the same. There is a segment of occasional computer users who would have serious trouble completing this process. You need both printing and scanning technology. This is not something you can do from your phone, at least not without some elite workarounds.

I’m soon greeted with a screen that references the vocabulary of the Amazon tracking information that tells me when my order of knockoff camera batteries has passed through Louisville.


But there is some fine print.


(You also might note two misspellings in the DoE warning.)

I don’t recall ever signing a waiver when I’ve gone to vote in person. I’ve also never read the Google Terms of Service, which probably are relevant, given how much information it likely knows about me. But should I worry about my vote being modified, suppressed, or simply seen by others?

As many in the cybersecurity and election worlds watch Alaska’s experiment unfold, Steve Friess in the Intercept in 2014 called our system “a security nightmare,” with vulnerabilities to hackers on both on the user’s machine and enroute to the servers prior to being counted.

Friess writes, “Computer scientists have already done some of these things in controlled laboratory experiments, in some cases attacking the same systems that Scytl has deployed in other jurisdictions around the world. In fact just this week Joseph Kiniry, a principal investigator at Galois, an international cybersecurity firm, asked his team to figure out ways to alter locked, supposedly un-editable PDFs remotely without detection. It took them, he said, a day.”

Scytl’s website says that their system “is the only solution in the market that truly protects the privacy of voters and the secrecy of intermediate results.” You can read far more from Scytl.

A neon-colored link appeared on the Division of Elections website this week touting the security systems in place (ostensibly after the Trump campaign has floated the idea that the election is rigged, but the linked document makes no mention of the electronic measures.

Sophie Kleeman wrote in Gizmodo this week about the lack of a verifiable paper trail for online ballots and highlighted the pardox of participating in our online democracy:

“Online voting is often viewed in the same way as online banking and shopping, but there’s a vital distinction: voting is based on anonymity. Banking and shopping are solidly linked to a person’s identity, and that identification is a key weapon for preventing fraud. Online voting, however, requires both anonymity and proof of identity,” wrote Kleeman. “It’s the height of irony, really: the internet, a bastion for the faceless and the nameless, gets in the way of total obscurity.”

When the time comes for our country to have a serious conversation about the security of our voting systems, I fear that digital cryptography will prove to be a clunky topic. Even the very simple online descriptions push my understating of tech and security. “… specialized masking protocol in the voter’s device that ensures that voters’ choices are never known by the online platform or by any third-party device…”

Quick, how would you explain a blockchain to your grandmother? What do you (or Chuck Grassely for that matter) know about SSL and encryption?

I joined the 955 Alaskans who as of Wednesday have completed their online ballot. More than 16,000 in the three days came in person to vote early or sent in ballots by mail. It took me a week to complete, but Alaskans who want to vote online can apply for a ballot as late as Monday, November, 7th.



Scytl maintain an active research and publishing effort, with freely-availalbe papers like “Implementation of a Leakage-Resilient ElGamal Key Encapsulation Mechanism” and Fine-Tuning Groth-Sahai Proofs (For the uninitiated, “Groth-Sahai proofs are efficient non-interactive zero-knowledge proofs that have found widespread use in pairing-based cryptography. We propose efficiency improvements of Groth-Sahai proofs in the SXDH setting, which is the one that yields the most efficient non-interactive zero-knowledge proofs.”)

I haven’t been able to get complete information on the cost of the online voting system yet. A FOIA’ed document available online that includes a review of proposals lists a $250,000 license fee, $125,000 in annual subscriptions subscriptions, and $14,394 for ballot processing each election(costing about $74 per returned vote in 2012 by those numbers, not counting any additional staff or DoE resources).The state’s online checkbook system showed a paymbent of $198,017 in FY2013. In my searches, SOE Software, a Scytl subsidiary, received payments of $4,499 in FY15, and $24,200 in FY16, but I do not have complete financial information.


I voted online so you don’t have to


Potential Alaska Pot Businesses Sprout Up


The state of Alaska has been accepting marijuana business applications since late February. Of the 200 applications that the state Marijuana Control Office has made public, there are of course, High North, The Great Alaska Kush Company, Musky Ox, Frozen Budz, and curiously, Subsistence Products.

It will still be a few months before the state approves any applications and the first shops open up to customers. But it’s clear now that legalized pot has drawn out entreprenuers from across the state. The early industry push looks to be centered in Anchorage, the Matanuska-Susitna Valley, and Fairbanks. But southeast Alaska and more remote locations like Nome and Tok have business at the ready.

It appears that more than anything, Alaskans want to grow.

Alaska Marijuana Initiated Applications as of 3-16-2016

More than 100 are hoping to be growers. But 36 applicants want to open retail stores, six seek to make concentrates, four would like to test marijuana.The state is publishing a list of applications, currently only available in a messy, multi-page PDF.

Data is from the state Marijuana Control Office and is current only as of 3/16/2016. Location data is from each public notice and may not be completely accurate. There is one Soldotna location in the Gulf of Alaska, for example.


Potential Alaska Pot Businesses Sprout Up


Alaska’s Deadly Skies Mapped


Alaska’s rural communities depend on a dedicated fleet of bush planes and pilots to move supplies and people across thousands of miles of roadless territory. Anyone who flies, drives, or boats know that there is risk inherent in getting around. No less than 321 people have died over the past 15 years in Alaska aviation crashes, according to a database maintained by the National Transportation Safety Board.

Click on each crash point for more information.

This map is better zooming and examining individual accidents. (Please don’t mind the web mercator projection…)

CartoDB’s quick and dirty heat map shows the obvious: Alaska’s urban core of Anchorage and the railbelt north towards the Mat-Su Valley is where many of the crashes occur. But all of bush Alaska relies on air transportation and freight for some purpose.

You also see a lot of activity in the southwest portion of the state, where the state’s third busiest airport, Bethel, serves as a hub for nearly 60 villages and in Southeast Alaska where the weather is (especially?) notoriously bad.


Flying into Tuluksak in November, 2014.

In 2015 alone, there have been 10 crashes that killed 22 people. To be honest, I had to stop and restart this project several times this fall when additional crashes changed the data set. A large majority of the crashes are limited to one or two deaths, but there are several with more than five fatalities.

Screen Shot 2015-11-29 at 10.49.56 PM

2015’s 10 crashes are not out of the ordinary: 10 of the past 15 years have seen 10 or more fatal crashes. On average, just under 22 people die annually in Alaskan aviation incidents. 2009 and 2014 each had just four fatal crashes, claiming seven and five lives respectively.

2013 was the most deadly year in recent memory, when 14 fatal crashes killed 35 people and seriously injured 12. One accident can  very quickly add up; 10 of the deaths came in a Rediske Air flight in Soldotna that is blamed on improper cargo storage.

I remember reporting for KYUK that fall on a Era Cessna 208 carrying six that went down near St. Mary’s in terrible weather Thanksgiving weekend. Four were killed that evening. Icing conditions may have played a role in that crash, but the NTSB has not come out with a final report.

Four died when a Cessna 208 crashed near St. Mary's. Photo courtesy of Alaska DPS.

Four died when a Cessna 208 crashed near St. Mary’s. Photo courtesy of Alaska DPS.


Screen Shot 2015-11-29 at 10.37.34 PMBut it’s not during Alaska’s cold, wet, and dark winters that tragedies claim the lives of pilots and passengers. The short three months of summer account for a disproportionate amount of crashes over the past 15 years. 55% of the fatal crashes in this period happened during June, July, and August. It’s a period of intense activity: the salmon are running, there’s no shortage of daylight. People naturally get around more this time of year and spend more time in the air.

On August 9th, 2010, a plane carrying former Senator Ted Stevens and eight others aboard a went down in the  Muklung Hills northeast of Dillingham. Even with a snow-free landscape, it still took hours for a rescue team to reach the hillside. Half of the plane’s occupants perished.


Former Senator Ted Stevens and four others died in while on a Bristol Bay fishing trip. Photo from Alaska DPS.

Former Senator Ted Stevens and four others died in while on a Bristol Bay fishing trip. Photo from Alaska DPS.

Winter is statistically a different story. February has only seen three fatal crashes in 15 years. July shows 29 crashes in the same period, killing 83 and seriously injuring six. August alone has seen more fatal crashes than the cold six-month stretch of November through April.

Screen Shot 2015-11-29 at 10.42.52 PM

Data for this visualization begin in 2001, when the NTSB began consistently publishing latitude and longitude data for crashes. You can further explore the data in this spreadsheet derived from a download of NTSB data.


This analysis looks at only crashes from 2001 to 2015 in the NTSB database that include a fatality; there are additional accidents and incidents in which the pilot and passengers survive or are uninjured. The NTSB is inconsistent in reporting several aspects of the dataset, such as make and model, purpose of flight, and FAR description. A few notes are included where latitude and longitude data had to be tweaked.

Data may be incomplete. Tools included CartoDB, Excel, Google Spreadsheets, and Plot.ly.


Alaska’s Deadly Skies Mapped


How Good Are Alaska’s Oracles At Predicting Oil Prices?

If I could predict crude oil prices, I wouldn’t be writing this, I’d be somewhere warm, writing a very different blog post. The Alaska Legislature finds itself in the same situation, except that they go someplace wet (Juneau) each year and build a budget in which (historically) around 90 percent of revenues come from the volatile commodity.

A couple months out from the 2016 session, we’re soon expecting the Alaska Department of Revenue’s Fall Revenue Sourcebook, one of the major tools we have for predicting oil prices and our future state budgets.

Screen Shot 2015-11-22 at 10.23.37 PM

We’re halfway through FY2016 now, which ends June 30, 2016. Let’s see how the experts DOR uses did over the last decade predicting this exact year.

FY 2016 Predicted Oil Price

The estimates range from a spartan $25.50  to a dreamy $114.88 per barrel of Alaska North Slope at West Coast refineries. While we still have another seven months before FY2016 is done, oil prices would have to climb in a hurry to break $60 for the year’s average.

Where do these numbers come from? Why do they exist? Why bother predicting more than a few months ahead? Maybe I should ask myself the same questions. It’s either that or don’t plan at all?

The DOR currently brings together experts in a closed-door session each fall for a day based on “a modified Delphi Method” to produce the oil price predictions that guide revenue models. Dermot Cole last year outlined a few of the quirks of the process developed by the RAND Corporation.

37 people from state government, business, and academia took part in 2014, according to the state. They each scribbled down their forecasts after “a day of presentations by experts on oil price markets and market structure.”

Can you picture eight hours of PowerPoint presentations and a lot of pondering and furrowed brows? I’m picturing pretty good catering?

After a look at the last decade of forecasts, time and global markets have not been kind to the prognosticators.  Digging through a trove of PDFs on the DOR website, I pulled forecasts going back to 2001.

Alaska West Coast Oil Price Prediction vs. Actual Price (1)

The predictions can be wildly off, especially more than a couple years out. The lean years of the late ’90s were not far out of memory when the prediction team anticipated oil staying low.

In November of 2005, when prices hovered in the mid-50s, DOR assumed a long-term average of $25.50 in the years to come (At that time, rules restricted long-term changes to once every two years upon agreement from forecasting participants.)

In October/November of 2013, the group predicted FY2016 oil to be near $107.69. (It was at 42.34 on November 19th). Triple digits had been the norm for some time by 2013. (The state-funded $24 million Bethel pool was under construction in 2013 and we were holding Knik Arm Bridge hearings.)

The forthcoming glut of new American and cheaply-produced Saudi oil had yet to be reflected in prices. It makes total sense now.

Few predicted the(ongoing) collapse in prices that began a year ago. I might actually have enough cash in my wallet right now to buy a barrel of oil. That is rarely true.

The 2014 estimate for FY2016 (which for short term forecasts relied on a new internal probabilistic model based on “Jump Diffusion”) called for more modest $66.03. While it sounds close, that’s still 34% above the current FY2016 daily average of 49.18. That’s a difference this year of $400+ million dollars in unrestricted reveue, according to 2014 assumptions.

The estimated prices in the above graph begin at most about 6-9 months after the time of the forecast in the first year in which participants have no actual data, that is, only true predictions are included. For example, the November 2001 group worked in middle of FY2002, looking forward to FY2003 which was to begin a half year ahead in July of 2002.

The following chart shows these initial  predictions of about seven to 18 months ahead.

 Most Recent Prediction Versus Actual Price.png

Until the recent plunge in prices, the oil market consistently outperformed expectations, including a near doubling in 2008. It gets worse if you try to predict further back.

Two years isn’t any better, and five years isn’t good for much.

Prediction from Five Years Prior Versus Actual Price (4).png

How large is the difference compared to the actual price? The Oracles consistently underestimated prices. Until they didn’t.

Actual Price Percentage Difference Compared to Forecasts (1).png

We would hope that these percentage difference lines would hover a bit closer to zero.

What’s to come? The 2015 Spring estimate looks forward to a surge. Oil hits $86.66 in FY2017, rises steadily to $109.54 by 2020, and is an attractive $124.34 in 2024. Never mind the forecast of just 320,300 barrels per day flowing down TAPS that year. Maybe we’ll have the first drops of LNG flowing by then.

Estimates are estimates. What happens in OPEC meetings, on the tar sands of Alberta, and within East Asian economies will ultimately determine the price. Right?

In any case, the state now makes more from appreciation and dividends from the $50 billion in the Permanent Fund than oil.

Dermot Cole said it best in his article:

“The future is uncertain.”

Alaska DOR said it second best on the cover of their cryptic 1989 Fall Revenue Sourcebook.

“Free Rides Die Hard.”

Screen Shot 2015-11-22 at 10.24.21 PM


Notes: Data come from Alaska Department of Revenue sourcebooks. Price data are from DOR forecast documents and the DOR Tax Division ‘s records. This used nominal values from fall forecasts with the exception of 2015,  which used the spring book.

I parsed data in Excel and charted in Plot.ly. Revenues to the state are based on the wellhead price, which subtracts costs and other items. I used ANS West Coast for consistency across tax regimes and to reflect DOR’s predictions on solely price instead of the many other factors it takes into account.

You can view data here.


Five Million Acres of Alaska Wildfire Growth

Alaska is on track to break 2004’s record for the most acres burned. Nearly 700 fires have been started by lightening storms or humans.

The Alaska Interagency Coordination Center serves as a clearinghouse for the many agencies that are responsible for tracking and fighting fires across the state and has a wealth of GIS data. This animation visualizes the Alaska Interagency Coordination Center’s measured perimeters for late June through July 18th.

Tools: QGIS 2.4, Time Manager, OpenLayers plugin, GoPro Studio.

Data: AICC, Google.

Note: Data are incomplete and do not perfectly correspond to fire activity on the display date. This is for experimental purposes only.


Aniak Complex Battles Massive Middle Kuskokwim Fires

Alaska is on track to break its all-time record for acres burned in a fire season. The southwest portion of the state has experienced the same dry conditions that caused the state to flare up with lightening storms rolled through in June. I had a chance to visit the “Aniak Complex” recently. As the state battles more than 300 fires, they’ve stationed crew leaders in remote areas to battle clusters of fires. I reported on the efforts for KYUK. It was largely an audio project, but I was able to get some interesting footage while traveling.