1 00:00:00,000 --> 00:00:12,840 We've got a bunch of brand new faces today, so I'll introduce myself and then we can do 2 00:00:12,840 --> 00:00:15,520 a round of introductions. 3 00:00:15,520 --> 00:00:17,680 So my name is Keegan. 4 00:00:17,680 --> 00:00:22,760 I've been in the cannabis space since the beginning of 2018. 5 00:00:22,760 --> 00:00:26,960 Started as a laboratory analyst at a lab that tests for cannabis. 6 00:00:26,960 --> 00:00:32,760 Then I started developing software and then I decided, you know what, it's time to just 7 00:00:32,760 --> 00:00:34,600 help everybody in the cannabis industry. 8 00:00:34,600 --> 00:00:40,640 So I started CanLinux to principally help laboratories, but really anyone who needs 9 00:00:40,640 --> 00:00:43,000 help with data. 10 00:00:43,000 --> 00:00:49,920 Now I would love to hear about what your interests are and what brings you to the group and what 11 00:00:49,920 --> 00:00:51,600 you may hope to learn. 12 00:00:51,600 --> 00:00:54,040 So I'll just start in my top left corner. 13 00:00:54,040 --> 00:00:58,080 Jerry, would you be interested in introducing yourself to the group? 14 00:00:58,080 --> 00:00:59,080 Sure. 15 00:00:59,080 --> 00:01:04,920 I'm a marketing communications and data analytics consultant. 16 00:01:04,920 --> 00:01:09,320 Just finished a very big project and don't have anything to do. 17 00:01:09,320 --> 00:01:17,640 So I want to do something, you know, get involved in doing an analytics project just to keep 18 00:01:17,640 --> 00:01:20,020 my hand in the game. 19 00:01:20,020 --> 00:01:23,760 And I saw the meetup and I said, wow, yeah, maybe I should do something with cannabis. 20 00:01:23,760 --> 00:01:27,400 And new and growing industry is probably something very interesting. 21 00:01:27,400 --> 00:01:30,720 Awesome to have you, Jerry. 22 00:01:30,720 --> 00:01:32,400 You're in the right place for certain. 23 00:01:32,400 --> 00:01:34,680 Yeah, I hope so. 24 00:01:34,680 --> 00:01:35,680 You are. 25 00:01:35,680 --> 00:01:38,320 You said, would you be interested in introducing yourself to the group? 26 00:01:38,320 --> 00:01:39,320 Hey, everyone. 27 00:01:39,320 --> 00:01:40,320 My name is Youssef. 28 00:01:40,320 --> 00:01:46,160 Actually, I work right now as a cloud solution consultant. 29 00:01:46,160 --> 00:01:51,600 And I discovered the meetup yesterday and told myself, why not try to meet people eager 30 00:01:51,600 --> 00:01:56,120 to learn like me, because I'm doing I'm trying to learn data science on the side on myself 31 00:01:56,120 --> 00:01:57,120 on my own. 32 00:01:57,120 --> 00:02:05,120 So yeah, it's a great opportunity to meet like minded people and ask for help and advice. 33 00:02:05,120 --> 00:02:06,120 Awesome. 34 00:02:06,120 --> 00:02:10,400 Great place to learn data science. 35 00:02:10,400 --> 00:02:15,040 Patrick, would you be interested in introducing yourself, please? 36 00:02:15,040 --> 00:02:16,040 Yeah, sure. 37 00:02:16,040 --> 00:02:17,040 Thanks. 38 00:02:17,040 --> 00:02:20,280 My name is Patrick Callahan in Wilmington, Delaware. 39 00:02:20,280 --> 00:02:22,760 It's 20 minutes outside of Philadelphia. 40 00:02:22,760 --> 00:02:27,960 And we have some clients in the cannabis space and we're a data science and analytics company. 41 00:02:27,960 --> 00:02:30,000 I compass read. 42 00:02:30,000 --> 00:02:31,000 Sorry. 43 00:02:31,000 --> 00:02:32,000 Awesome. 44 00:02:32,000 --> 00:02:33,000 Awesome. 45 00:02:33,000 --> 00:02:35,000 Delaware is on the agenda for today. 46 00:02:35,000 --> 00:02:37,480 So in fact, we're going to be talking about each and every state. 47 00:02:37,480 --> 00:02:38,480 So stay tuned. 48 00:02:38,480 --> 00:02:40,880 It's going to be an exciting day. 49 00:02:40,880 --> 00:02:45,200 Michelle, would you be interested in introducing yourself? 50 00:02:45,200 --> 00:02:47,200 Hello. 51 00:02:47,200 --> 00:02:52,680 Like some people said, I saw the meetup and I was like, this is a yes for me. 52 00:02:52,680 --> 00:02:58,520 There's a couple of companies that I follow to get information on a couple of doctors 53 00:02:58,520 --> 00:03:03,600 who have companies in other states who have been practicing medicinal use of marijuana 54 00:03:03,600 --> 00:03:05,440 for like a decade. 55 00:03:05,440 --> 00:03:07,760 So I'm basically just trying to get as much exposure as I can. 56 00:03:07,760 --> 00:03:13,680 I started my own cannabis coaching company last year to help people and teach people 57 00:03:13,680 --> 00:03:20,600 how to microdose cannabis for ADHD, anxiety, depression. 58 00:03:20,600 --> 00:03:23,560 I'm autistic, so it helps with that. 59 00:03:23,560 --> 00:03:27,040 Also, trauma, yada, yada, yada. 60 00:03:27,040 --> 00:03:33,680 And I created my own dosing system and it's an oral delivery and very easy to dose. 61 00:03:33,680 --> 00:03:35,480 It's not by milligrams, it's by ounces. 62 00:03:35,480 --> 00:03:37,280 So it's by volume. 63 00:03:37,280 --> 00:03:44,400 But anyway, it's helping people be able to microdose it without having to deal with being 64 00:03:44,400 --> 00:03:45,680 like affected and stuff. 65 00:03:45,680 --> 00:03:50,320 So I'm filling in gaps with science in this way also. 66 00:03:50,320 --> 00:03:51,760 Awesome. 67 00:03:51,760 --> 00:03:55,160 Your expertise is going to come in real handy in some future meetups. 68 00:03:55,160 --> 00:04:00,720 So we're going to be diving a lot into the production and consumption of cannabis here 69 00:04:00,720 --> 00:04:01,720 in 2022. 70 00:04:01,720 --> 00:04:05,500 So today we'll be wrapping up a sales forecast. 71 00:04:05,500 --> 00:04:10,600 So in future meetups, as well as today, your expertise will be valuable. 72 00:04:10,600 --> 00:04:12,600 Happy to help. 73 00:04:12,600 --> 00:04:13,600 Awesome. 74 00:04:13,600 --> 00:04:18,880 My board changed a little bit here, but Devon, would you be interested in introducing yourself? 75 00:04:18,880 --> 00:04:19,880 Hi. 76 00:04:19,880 --> 00:04:24,000 Yes, just working on my camera now. 77 00:04:24,000 --> 00:04:25,000 Devon Moftano. 78 00:04:25,000 --> 00:04:29,120 I'm working in Wilmington, Delaware at Compass Red. 79 00:04:29,120 --> 00:04:30,120 Awesome. 80 00:04:30,120 --> 00:04:36,920 Raul, would you be interested in introducing yourself, please? 81 00:04:36,920 --> 00:04:37,920 Yes. 82 00:04:37,920 --> 00:04:38,920 Hi. 83 00:04:38,920 --> 00:04:39,920 Good morning. 84 00:04:39,920 --> 00:04:40,920 My name is Raul Infante. 85 00:04:40,920 --> 00:04:50,320 I am a medical marijuana patient in Connecticut, and I am also a grower as Connecticut. 86 00:04:50,320 --> 00:04:53,540 Ever since October, we've been able to grow our own cannabis. 87 00:04:53,540 --> 00:05:00,400 I have volunteered and I have been an employee at multiple hemp farms ever since the year 88 00:05:00,400 --> 00:05:05,280 2019, helping them with the harvest and helping them with the seeding and with the taking 89 00:05:05,280 --> 00:05:07,980 care of the plant through vegetation stage. 90 00:05:07,980 --> 00:05:13,480 I got a few grows under my belt, even though the patient program just started in October. 91 00:05:13,480 --> 00:05:16,040 I've got a few grows. 92 00:05:16,040 --> 00:05:21,520 I've been growing since about 2017 in Massachusetts with my cousin for himself. 93 00:05:21,520 --> 00:05:26,400 I've shown and helped him how to grow. 94 00:05:26,400 --> 00:05:36,120 Right now, I am currently the head intern at a farm in Danbury called Seymourone Farmstead, 95 00:05:36,120 --> 00:05:40,100 S-E-A-M-A-R-R-O-N Farmstead. 96 00:05:40,100 --> 00:05:45,040 We will be growing hemp this upcoming spring season. 97 00:05:45,040 --> 00:05:47,440 That's just a little bit of background of where I'm coming from. 98 00:05:47,440 --> 00:05:52,920 I'll be starting horticulture classes in June in one of our local community colleges here 99 00:05:52,920 --> 00:05:55,600 in Southern Connecticut. 100 00:05:55,600 --> 00:05:56,600 That's just a little bit of background. 101 00:05:56,600 --> 00:05:57,600 I'm a father. 102 00:05:57,600 --> 00:05:58,600 I have a four-year-old. 103 00:05:58,600 --> 00:06:05,320 Yeah, things are pretty high energy at my crib, at the house. 104 00:06:05,320 --> 00:06:08,760 Thank you for having me and I appreciate you very much. 105 00:06:08,760 --> 00:06:09,760 Thank you for joining us. 106 00:06:09,760 --> 00:06:12,000 Love to have your expertise as well. 107 00:06:12,000 --> 00:06:14,000 We have a lot of data scientists. 108 00:06:14,000 --> 00:06:17,680 Sorry, I didn't mean to interrupt, sir. 109 00:06:17,680 --> 00:06:21,040 I'm probably just going to have you in my ear because I'm at work right now and I just 110 00:06:21,040 --> 00:06:23,080 wanted to be a part of it. 111 00:06:23,080 --> 00:06:24,080 Definitely. 112 00:06:24,080 --> 00:06:25,080 Definitely. 113 00:06:25,080 --> 00:06:27,840 You can listen in and then watch the recording later. 114 00:06:27,840 --> 00:06:33,720 Always happy to have your expertise too because it's so important to hear from the producers 115 00:06:33,720 --> 00:06:38,920 and processors and retailers and everybody in the industry. 116 00:06:38,920 --> 00:06:42,720 Really interested about the Northeast because I don't know as much about the industry. 117 00:06:42,720 --> 00:06:46,040 What part of the country are you in? 118 00:06:46,040 --> 00:06:47,040 California, sir? 119 00:06:47,040 --> 00:06:49,080 What's your name? 120 00:06:49,080 --> 00:06:53,040 Well, Kenlytics is based out of Olympia, Washington. 121 00:06:53,040 --> 00:06:55,880 Right now I'm visiting family in North Carolina. 122 00:06:55,880 --> 00:07:02,960 So worked with people all over the country, worked with a couple of people in Massachusetts, 123 00:07:02,960 --> 00:07:05,640 but I only know a cursory amount. 124 00:07:05,640 --> 00:07:07,480 So always happy to learn myself. 125 00:07:07,480 --> 00:07:08,480 Yeah, me too. 126 00:07:08,480 --> 00:07:09,480 Always. 127 00:07:09,480 --> 00:07:10,480 Me too. 128 00:07:10,480 --> 00:07:11,480 What's your name, sir? 129 00:07:11,480 --> 00:07:12,480 What can I call you? 130 00:07:12,480 --> 00:07:13,480 Kenlytics. 131 00:07:13,480 --> 00:07:14,480 What was that? 132 00:07:14,480 --> 00:07:15,480 Exactly. 133 00:07:15,480 --> 00:07:20,000 And I got into this space helping out laboratories and my background's in data science. 134 00:07:20,000 --> 00:07:22,680 So just trying to combine the two. 135 00:07:22,680 --> 00:07:23,680 What's your name? 136 00:07:23,680 --> 00:07:24,680 Megan. 137 00:07:24,680 --> 00:07:25,680 Megan? 138 00:07:25,680 --> 00:07:26,680 Yes. 139 00:07:26,680 --> 00:07:27,680 Megan? 140 00:07:27,680 --> 00:07:28,680 All right, Megan. 141 00:07:28,680 --> 00:07:29,680 Nice to meet you, sir. 142 00:07:29,680 --> 00:07:30,680 Good to meet you, Ronald. 143 00:07:30,680 --> 00:07:31,680 I'll be I'm gonna be on mute. 144 00:07:31,680 --> 00:07:32,680 All right. 145 00:07:32,680 --> 00:07:33,680 Okay. 146 00:07:33,680 --> 00:07:34,680 Thank you. 147 00:07:34,680 --> 00:07:36,680 David, would you be introduced? 148 00:07:36,680 --> 00:07:39,600 I need to think of something else to say. 149 00:07:39,600 --> 00:07:41,600 David, would you mind introducing yourself? 150 00:07:41,600 --> 00:07:42,600 Sure. 151 00:07:42,600 --> 00:07:44,760 So hello, everybody from Denver. 152 00:07:44,760 --> 00:07:51,760 So I work for the state of Colorado in marijuana enforcement division, and I am a data scientist. 153 00:07:51,760 --> 00:07:54,680 So I'm happy to be here. 154 00:07:54,680 --> 00:07:58,200 Awesome to have you here, especially from the state of Colorado. 155 00:07:58,200 --> 00:08:01,560 So love that you're joining. 156 00:08:01,560 --> 00:08:07,600 So hopefully you can feel free to share and then maybe you could learn a thing or two. 157 00:08:07,600 --> 00:08:09,600 So it's awesome to have you. 158 00:08:09,600 --> 00:08:12,280 So we're just here to learn from each other. 159 00:08:12,280 --> 00:08:13,280 Thanks. 160 00:08:13,280 --> 00:08:14,280 Awesome. 161 00:08:14,280 --> 00:08:16,680 So got four other people here. 162 00:08:16,680 --> 00:08:19,280 There's Kelvin, it looks like. 163 00:08:19,280 --> 00:08:22,280 Would you mind introducing yourself? 164 00:08:22,280 --> 00:08:28,680 Okay, let's see. 165 00:08:28,680 --> 00:08:36,600 Okay, we may come back to you, Kelvin. 166 00:08:36,600 --> 00:08:40,520 Matthew, would you mind introducing yourself? 167 00:08:40,520 --> 00:08:41,520 Yes, absolutely. 168 00:08:41,520 --> 00:08:44,520 So I'm a software developer. 169 00:08:44,520 --> 00:08:48,240 I'm actually one of the people also. 170 00:08:48,240 --> 00:08:54,520 Looks like we're about half the group here, which is kind of cool. 171 00:08:54,520 --> 00:08:57,240 But yeah, just this is kind of a unique meetup. 172 00:08:57,240 --> 00:09:01,280 And so we wanted to be here since we heard about it. 173 00:09:01,280 --> 00:09:03,280 So thanks for putting this on. 174 00:09:03,280 --> 00:09:04,520 I appreciate it. 175 00:09:04,520 --> 00:09:05,520 Thanks for joining. 176 00:09:05,520 --> 00:09:10,960 And I always say to everybody, there's a shortage of data scientists and software developers. 177 00:09:10,960 --> 00:09:15,120 So your value you can bring will go a long way. 178 00:09:15,120 --> 00:09:16,120 Thank you. 179 00:09:16,120 --> 00:09:19,120 So we've got Michelle. 180 00:09:19,120 --> 00:09:21,120 Let's see. 181 00:09:21,120 --> 00:09:24,040 Ben, would you be introduced? 182 00:09:24,040 --> 00:09:25,920 Would you like to introduce yourself? 183 00:09:25,920 --> 00:09:26,920 Yeah. 184 00:09:26,920 --> 00:09:27,920 Hi. 185 00:09:27,920 --> 00:09:35,160 Also from the Compass Red team here in Philadelphia, data analysts and product developer. 186 00:09:35,160 --> 00:09:41,560 On the team, like was mentioned, we have some clients in the cannabis space. 187 00:09:41,560 --> 00:09:47,280 Our team primarily works in R and R Shiny, as well as all of the dashboarding applications 188 00:09:47,280 --> 00:09:49,920 under the sun. 189 00:09:49,920 --> 00:09:57,720 The focus of my work has primarily been in data visualization in the R world as well 190 00:09:57,720 --> 00:09:58,720 as D3. 191 00:09:58,720 --> 00:09:59,720 So happy to be here. 192 00:09:59,720 --> 00:10:00,720 Awesome to have you. 193 00:10:00,720 --> 00:10:05,480 And we've had a lot of people who've worked with R in the past, and I always think it's 194 00:10:05,480 --> 00:10:10,280 just a programming languages are a tool, a means to an end. 195 00:10:10,280 --> 00:10:13,040 So I do a lot of work in Python. 196 00:10:13,040 --> 00:10:15,680 So I've got a bit of code to show you today in Python. 197 00:10:15,680 --> 00:10:21,040 However, it'd be awesome if you were ambitious enough to rewrite it in R. So that would go 198 00:10:21,040 --> 00:10:24,040 a long way too. 199 00:10:24,040 --> 00:10:26,880 All right. 200 00:10:26,880 --> 00:10:32,440 I may have butchered the pronunciation of that. 201 00:10:32,440 --> 00:10:36,680 Would you mind introducing yourself if you would like? 202 00:10:36,680 --> 00:10:40,080 OK. 203 00:10:40,080 --> 00:10:47,600 So like D or Kelvin, you're welcome to speak up at any point if you want to introduce yourself. 204 00:10:47,600 --> 00:10:48,600 You don't have to. 205 00:10:48,600 --> 00:10:51,320 So options always there on the team. 206 00:10:51,320 --> 00:10:52,520 All right. 207 00:10:52,520 --> 00:10:57,360 Well, for those of you who are new to the group, I don't really like to steal the show. 208 00:10:57,360 --> 00:10:59,360 It just ends up happening. 209 00:10:59,360 --> 00:11:02,800 It is a meet up here after all. 210 00:11:02,800 --> 00:11:08,640 But long story short, I usually prepare a script and a short presentation just to sort 211 00:11:08,640 --> 00:11:11,120 of give the group a direction. 212 00:11:11,120 --> 00:11:14,120 And then usually that takes around 30 minutes or so. 213 00:11:14,120 --> 00:11:17,720 And then have a 15 minute or so discussion at the end. 214 00:11:17,720 --> 00:11:20,660 However, we've got a huge group today. 215 00:11:20,660 --> 00:11:26,280 So if you want to chime in at any point, just start talking up. 216 00:11:26,280 --> 00:11:31,040 And we can start a dialogue at any point. 217 00:11:31,040 --> 00:11:39,040 But without further ado, I'll just go ahead and share my screen just so you can follow 218 00:11:39,040 --> 00:11:40,660 along. 219 00:11:40,660 --> 00:11:46,760 So for those of you who are new, everything's available on GitHub. 220 00:11:46,760 --> 00:11:51,240 So you can check out all the work we've done in 2021. 221 00:11:51,240 --> 00:11:55,560 And then I try to give more than 20 minutes. 222 00:11:55,560 --> 00:12:04,100 But I try to get the script for the day uploaded beforehand so that way you can follow along. 223 00:12:04,100 --> 00:12:13,280 And then just put together just a short presentation just to start talking about the subject for 224 00:12:13,280 --> 00:12:14,280 the day. 225 00:12:14,280 --> 00:12:20,200 And so last week, we started talking about, OK, what is data science and where did it 226 00:12:20,200 --> 00:12:22,280 come from? 227 00:12:22,280 --> 00:12:27,240 And I said, OK, it's a mixture of statistics and computer science. 228 00:12:27,240 --> 00:12:29,400 And my background is in economics. 229 00:12:29,400 --> 00:12:32,720 So I like to splash in a little bit of economics. 230 00:12:32,720 --> 00:12:35,800 But really, it just picks from a lot of different fields. 231 00:12:35,800 --> 00:12:37,800 So I did a little research. 232 00:12:37,800 --> 00:12:46,400 And it looks like, OK, data science has been around 40 plus years now. 233 00:12:46,400 --> 00:12:54,160 So we've got Peter Noir, who didn't like the term computer science. 234 00:12:54,160 --> 00:12:58,300 And he thought it would be better called data science. 235 00:12:58,300 --> 00:12:59,420 So that's interesting. 236 00:12:59,420 --> 00:13:07,760 And then you've got CF Jeff Wu, who is an engineering statistician at the Georgia Institute 237 00:13:07,760 --> 00:13:08,760 for Technology. 238 00:13:08,760 --> 00:13:16,760 And he was the first one to use data science as an alternative name for statistics. 239 00:13:16,760 --> 00:13:20,360 So just a little bit of history. 240 00:13:20,360 --> 00:13:27,160 It's always kind of interesting to know where these things came about from. 241 00:13:27,160 --> 00:13:36,760 Latonya Sweeney, she was the chief technologist at the Federal Trade Commission, I believe. 242 00:13:36,760 --> 00:13:37,760 For a period. 243 00:13:37,760 --> 00:13:41,760 And I believe she's currently a Harvard professor. 244 00:13:41,760 --> 00:13:48,760 So she's probably forgotten more than I know. 245 00:13:48,760 --> 00:13:50,760 And she's just an expert here. 246 00:13:50,760 --> 00:13:55,760 And so she said, OK, there's three trends to watch for in data science. 247 00:13:55,760 --> 00:13:58,760 One, we're just going to be collecting more data. 248 00:13:58,760 --> 00:14:06,480 Two, the data is going to be getting more granular, to the point where it's sort of 249 00:14:06,480 --> 00:14:09,560 like collected if you can. 250 00:14:09,560 --> 00:14:12,160 The data is getting so granular. 251 00:14:12,160 --> 00:14:20,560 And there's so much of it that we've now reached what some are calling the data deluge, which 252 00:14:20,560 --> 00:14:25,920 is there's just, I like to call it a fire hose of data. 253 00:14:25,920 --> 00:14:31,560 So think of a fire hydrant that's bust a cap, and it's just spraying out data. 254 00:14:31,560 --> 00:14:35,240 And so first things first, you have to wrangle it. 255 00:14:35,240 --> 00:14:37,880 And then collect it. 256 00:14:37,880 --> 00:14:40,960 And then think about what you're going to do with it. 257 00:14:40,960 --> 00:14:44,520 So essentially, you need to design your data. 258 00:14:44,520 --> 00:14:46,240 And you need to collect it. 259 00:14:46,240 --> 00:14:49,960 Then you can't forget about analyzing your data. 260 00:14:49,960 --> 00:14:54,560 So there's a little bit of data science for today. 261 00:14:54,560 --> 00:15:01,120 And now what we're all here for is about the cannabis industry. 262 00:15:01,120 --> 00:15:08,360 So as a good Bayesian, I thought I would ask some of you what your priors may be. 263 00:15:08,360 --> 00:15:15,040 So here are three topics that we'll be talking about today. 264 00:15:15,040 --> 00:15:19,380 And we can make forecasts and check our forecasts. 265 00:15:19,380 --> 00:15:27,500 But before we begin, does anybody have any expectations of what states may permit medicine 266 00:15:27,500 --> 00:15:30,720 or adult use in 2022? 267 00:15:30,720 --> 00:15:38,680 And what you may expect sales to be in, say, New York or just in the US in general? 268 00:15:38,680 --> 00:15:39,680 Anyone? 269 00:15:39,680 --> 00:15:41,680 Feel free to... 270 00:15:41,680 --> 00:15:45,560 I'll speak on New York. 271 00:15:45,560 --> 00:15:46,560 I'm in Brooklyn. 272 00:15:46,560 --> 00:15:47,560 Yes. 273 00:15:47,560 --> 00:15:49,240 I'm from Western New York. 274 00:15:49,240 --> 00:15:51,840 So I'm just in this area in general. 275 00:15:51,840 --> 00:15:55,440 And it obviously recently became legal here. 276 00:15:55,440 --> 00:16:04,320 I've seen some head shops go up and come down really quickly, like block by block. 277 00:16:04,320 --> 00:16:05,320 There was so many. 278 00:16:05,320 --> 00:16:06,760 But you needed a card. 279 00:16:06,760 --> 00:16:10,800 So now that you don't need a card, it's a little different. 280 00:16:10,800 --> 00:16:16,920 But every single deli here sells in some form. 281 00:16:16,920 --> 00:16:21,500 Like liquid, like there's beverages. 282 00:16:21,500 --> 00:16:23,920 You can literally get it on every single corner. 283 00:16:23,920 --> 00:16:25,760 And it's not like weed for sale here. 284 00:16:25,760 --> 00:16:30,120 It's like we have this, but we call it this. 285 00:16:30,120 --> 00:16:34,960 But you can literally buy weed from any single deli that sells it. 286 00:16:34,960 --> 00:16:36,520 And then there's the other things. 287 00:16:36,520 --> 00:16:42,920 I don't know if people have heard of like kratom and all the other plants that are alternatives 288 00:16:42,920 --> 00:16:43,920 to marijuana. 289 00:16:43,920 --> 00:16:45,360 So there's a lot of that going on. 290 00:16:45,360 --> 00:16:48,760 So I feel like for me, it's really chaotic. 291 00:16:48,760 --> 00:16:52,680 And it's not something I pay too much attention to because my focus is more... 292 00:16:52,680 --> 00:16:54,160 My target is... 293 00:16:54,160 --> 00:16:57,680 I also have a marketing background and I'm a hairstylist of 22 years. 294 00:16:57,680 --> 00:17:00,280 So I have a chemistry brain in that field. 295 00:17:00,280 --> 00:17:12,440 But my target is mostly middle-aged women, two or three kids deep, or stoners. 296 00:17:12,440 --> 00:17:15,520 There's like no in between. 297 00:17:15,520 --> 00:17:23,520 But for me here, it's very chaotic because there's a lot of dealers here that are half 298 00:17:23,520 --> 00:17:24,840 legal and half not legal. 299 00:17:24,840 --> 00:17:28,120 So they'll do events out in the streets and they won't need any permits because they can 300 00:17:28,120 --> 00:17:30,400 just go before they get in trouble. 301 00:17:30,400 --> 00:17:33,400 But everyone loves it and nobody cares. 302 00:17:33,400 --> 00:17:34,400 Cops don't care about it. 303 00:17:34,400 --> 00:17:35,760 It's just a very bizarre vibe. 304 00:17:35,760 --> 00:17:41,080 So I can't even answer any of these questions because it still feels like chaos. 305 00:17:41,080 --> 00:17:45,160 And I feel like it's going to be like that here, at least in New York City for a while, 306 00:17:45,160 --> 00:17:46,160 like years. 307 00:17:46,160 --> 00:17:47,160 Yes. 308 00:17:47,160 --> 00:17:54,800 I'm also wondering in New York State, just passed a deadline for individual municipalities 309 00:17:54,800 --> 00:18:05,560 to opt in or opt out for allowing retail sales and consumption venues for recreational marijuana. 310 00:18:05,560 --> 00:18:07,120 And I don't know the results on that yet. 311 00:18:07,120 --> 00:18:08,120 That would be interesting. 312 00:18:08,120 --> 00:18:09,840 I know my... 313 00:18:09,840 --> 00:18:12,840 I live in a village within a town. 314 00:18:12,840 --> 00:18:14,080 The village voted yes. 315 00:18:14,080 --> 00:18:15,320 The town voted no. 316 00:18:15,320 --> 00:18:17,760 They voted to opt out. 317 00:18:17,760 --> 00:18:19,800 So talking about chaos. 318 00:18:19,800 --> 00:18:26,680 Then there's also the question, I had a client a while ago who makes non-consumable hemp 319 00:18:26,680 --> 00:18:27,680 products. 320 00:18:27,680 --> 00:18:31,040 He has a cannabis farm up here. 321 00:18:31,040 --> 00:18:35,920 So I'm just wondering how that relates because there are so many products that you can get 322 00:18:35,920 --> 00:18:41,600 from hemp that are not psychotropic. 323 00:18:41,600 --> 00:18:42,600 Is that the right word? 324 00:18:42,600 --> 00:18:45,000 I think so. 325 00:18:45,000 --> 00:18:46,000 Yes. 326 00:18:46,000 --> 00:18:53,520 Well, after listening to you two, I may have to update my priors. 327 00:18:53,520 --> 00:18:55,080 That's real informative. 328 00:18:55,080 --> 00:18:58,400 So it's going to be interesting to see... 329 00:18:58,400 --> 00:19:01,480 Let's just make sure this... 330 00:19:01,480 --> 00:19:12,520 It's going to be interesting to see, one, if the gray market in New York becomes legalized, 331 00:19:12,520 --> 00:19:15,840 by the end of 2022. 332 00:19:15,840 --> 00:19:25,480 And then as far as the towns or locales opting out, from the research I've done, the demand 333 00:19:25,480 --> 00:19:29,760 for cannabis appears to be fairly inelastic. 334 00:19:29,760 --> 00:19:35,120 So that would suggest that people will just drive to the next town over and just bear 335 00:19:35,120 --> 00:19:36,120 the cost. 336 00:19:36,120 --> 00:19:39,640 But only so many people will do that. 337 00:19:39,640 --> 00:19:44,120 And it still may curtail consumption a little bit. 338 00:19:44,120 --> 00:19:50,000 So that may curtail sales to a certain extent. 339 00:19:50,000 --> 00:19:53,640 Hard to say. 340 00:19:53,640 --> 00:19:55,140 Awesome predictions. 341 00:19:55,140 --> 00:20:01,280 So for me, my Bayesian prior was, I'm not certain... 342 00:20:01,280 --> 00:20:05,920 Just to be honest, I'm not certain New York was going to get online in 2022. 343 00:20:05,920 --> 00:20:11,960 But after hearing from you two, it sounds like maybe the ball is moving a bit quicker 344 00:20:11,960 --> 00:20:14,840 than I anticipated there. 345 00:20:14,840 --> 00:20:28,920 And then for the US sales, I just had really no idea what a good prior would be. 346 00:20:28,920 --> 00:20:34,640 So not the best prior, but basically if I had to give one, I would say between like 347 00:20:34,640 --> 00:20:38,680 15 and 60 billion. 348 00:20:38,680 --> 00:20:43,440 But we'll get to that in just one second. 349 00:20:43,440 --> 00:20:48,520 But some states to look out for. 350 00:20:48,520 --> 00:20:54,280 So just want to let you know that I got this from the Cannabis Business Times. 351 00:20:54,280 --> 00:21:00,840 They published an article recently, 15 states that could legalize in 2022. 352 00:21:00,840 --> 00:21:02,560 And here they are. 353 00:21:02,560 --> 00:21:07,800 There could be some wild cards that just come that you don't expect. 354 00:21:07,800 --> 00:21:09,480 That's happened before. 355 00:21:09,480 --> 00:21:13,360 But these are the ones that they said to watch out for. 356 00:21:13,360 --> 00:21:18,520 So of course, states like New York, they've already permitted adult use. 357 00:21:18,520 --> 00:21:24,720 We're just waiting for the rollout similar in New Jersey and so on and so forth. 358 00:21:24,720 --> 00:21:30,480 But just starting from the top, Delaware, Oklahoma, Mississippi, Maryland. 359 00:21:30,480 --> 00:21:35,880 I think maybe one of the more likely ones. 360 00:21:35,880 --> 00:21:36,880 Same for Oklahoma. 361 00:21:36,880 --> 00:21:44,720 I would, Oklahoma has such a permissible market that it's essentially already adult use. 362 00:21:44,720 --> 00:21:47,280 You still technically have to be a patient. 363 00:21:47,280 --> 00:21:53,120 So I wouldn't be surprised if they formally allow adult use in Oklahoma. 364 00:21:53,120 --> 00:21:58,360 Ohio, things are trying to move along there with signatures. 365 00:21:58,360 --> 00:22:00,320 Same with Wyoming. 366 00:22:00,320 --> 00:22:03,280 There's efforts in Pennsylvania. 367 00:22:03,280 --> 00:22:08,520 Rhode Island looks like another one that I would say is a bit more likely. 368 00:22:08,520 --> 00:22:11,840 So I think they're waiting on a special session there. 369 00:22:11,840 --> 00:22:18,200 But basically, a lot of trying to think about Rhode Island's geography. 370 00:22:18,200 --> 00:22:24,040 But you've got states nearby that are all allowing adult use. 371 00:22:24,040 --> 00:22:28,360 So right in that northeast, you've got Massachusetts. 372 00:22:28,360 --> 00:22:32,760 Connecticut is moving the ball forward, if not already has. 373 00:22:32,760 --> 00:22:35,000 Then of course, New York and New Jersey. 374 00:22:35,000 --> 00:22:37,960 So I think that'll put pressure on Rhode Island. 375 00:22:37,960 --> 00:22:41,440 And so I wouldn't be surprised if it moves quickly. 376 00:22:41,440 --> 00:22:45,160 Nebraska may allow medicinal. 377 00:22:45,160 --> 00:22:51,440 Arkansas is working on adult use, or advocates are working on it. 378 00:22:51,440 --> 00:22:59,520 Florida, people are trying to get adult use onto the 2022 ballot. 379 00:22:59,520 --> 00:23:03,120 Same for Missouri. 380 00:23:03,120 --> 00:23:05,320 Just going to allow some people in. 381 00:23:05,320 --> 00:23:08,840 In all of the newcomers, I'll try to give you a chance to introduce yourselves 382 00:23:08,840 --> 00:23:12,480 and talk here at the end. 383 00:23:12,480 --> 00:23:15,080 North Carolina, I was surprised about this. 384 00:23:15,080 --> 00:23:18,880 But there's some efforts there for medicinal cannabis. 385 00:23:18,880 --> 00:23:22,800 But moving quite slowly, that would be expected. 386 00:23:22,800 --> 00:23:25,800 And then South Dakota, an interesting one. 387 00:23:25,800 --> 00:23:29,200 They're trying to get adult use on the 2022 ballot. 388 00:23:29,200 --> 00:23:32,440 And Idaho. 389 00:23:32,440 --> 00:23:35,840 Idaho is an interesting one because you've 390 00:23:35,840 --> 00:23:39,840 got some neighbors like Washington and Oregon 391 00:23:39,840 --> 00:23:44,280 who have had adult use cannabis for a long time now. 392 00:23:44,280 --> 00:23:49,360 And so Idaho is thinking about getting medicinal. 393 00:23:49,360 --> 00:23:52,320 So these are some states that are coming along. 394 00:23:52,320 --> 00:24:01,640 And the reason why these matter is I saw CNBC put out a forecast for, 395 00:24:01,640 --> 00:24:02,760 they beat me to it. 396 00:24:02,760 --> 00:24:07,400 So they put out a forecast for 2022. 397 00:24:07,400 --> 00:24:12,280 And so as a Bayesian, I have to incorporate this into my prior. 398 00:24:12,280 --> 00:24:15,920 So I would have liked to had it made my forecast before seeing this. 399 00:24:15,920 --> 00:24:19,040 But oh, well, I saw these numbers. 400 00:24:19,040 --> 00:24:23,600 So CNBC is predicting total sales in the US 401 00:24:23,600 --> 00:24:27,800 is going to be $31 billion, up 28%. 402 00:24:27,800 --> 00:24:30,640 And that large growth is probably going 403 00:24:30,640 --> 00:24:33,680 to be from potentially new states coming online or things 404 00:24:33,680 --> 00:24:36,160 quickening in certain states. 405 00:24:36,160 --> 00:24:40,960 And they're estimating, it was hard to tell from the audio clip, 406 00:24:40,960 --> 00:24:43,680 but it sounded like they were estimating New York 407 00:24:43,680 --> 00:24:47,320 to hit $4 billion in 2022. 408 00:24:47,320 --> 00:24:50,640 Personally, I don't know how realistic that would be, 409 00:24:50,640 --> 00:24:53,280 but never rule anything out. 410 00:24:53,280 --> 00:24:57,160 So New York is a big populous state. 411 00:24:57,160 --> 00:25:01,280 So it may be possible. 412 00:25:01,280 --> 00:25:04,120 Question, comment? 413 00:25:04,120 --> 00:25:07,120 New York is eager to make money off of it for sure. 414 00:25:07,120 --> 00:25:09,160 So it doesn't surprise me the speed at which they 415 00:25:09,160 --> 00:25:11,160 think it's going to grow, because especially here 416 00:25:11,160 --> 00:25:14,200 in New York City, one of the other things that I was going to say 417 00:25:14,200 --> 00:25:15,840 was there's a lot of fraud here. 418 00:25:15,840 --> 00:25:17,640 So when CBD became a thing, and you 419 00:25:17,640 --> 00:25:20,760 could get that at every deli, then there 420 00:25:20,760 --> 00:25:23,440 was a realization about how much of that product 421 00:25:23,440 --> 00:25:27,800 was just being sold willy nilly, and it's not right. 422 00:25:27,800 --> 00:25:29,360 So I feel like the same thing is going 423 00:25:29,360 --> 00:25:31,920 to happen with cannabis, especially with carts 424 00:25:31,920 --> 00:25:34,480 and stuff like that. 425 00:25:34,480 --> 00:25:37,320 So I think that New York doesn't care. 426 00:25:37,320 --> 00:25:38,960 They're just interested in the revenue. 427 00:25:38,960 --> 00:25:40,720 So the faster everything can roll out. 428 00:25:40,720 --> 00:25:44,440 So maybe that's why the prediction is so large. 429 00:25:44,440 --> 00:25:45,360 You think? 430 00:25:45,360 --> 00:25:48,520 Yes, and like you said, it sounds like people are already 431 00:25:48,520 --> 00:25:49,320 set up. 432 00:25:49,320 --> 00:25:51,920 So the infrastructure may be in place. 433 00:25:51,920 --> 00:25:55,680 So once they permit it, it could go quickly. 434 00:25:55,680 --> 00:25:58,520 And I've said this in prior meetups, 435 00:25:58,520 --> 00:26:03,280 but I'd like to say states, there's always 436 00:26:03,280 --> 00:26:04,360 an opportunity choice. 437 00:26:04,360 --> 00:26:07,280 When you choose to do something or not to do something, 438 00:26:07,280 --> 00:26:08,880 there's an opportunity cost. 439 00:26:08,880 --> 00:26:12,480 And so what I basically am saying is like, hey, states, 440 00:26:12,480 --> 00:26:15,200 if you're not permitting this, then like you said, 441 00:26:15,200 --> 00:26:18,000 there is this, it's not even gray, 442 00:26:18,000 --> 00:26:23,320 it's a black market going on that you're not taxing. 443 00:26:23,320 --> 00:26:26,040 And so the tax revenue could potentially 444 00:26:26,040 --> 00:26:29,640 go to socially beneficial things. 445 00:26:29,640 --> 00:26:32,640 I always use the example of building schools 446 00:26:32,640 --> 00:26:36,280 because you can put a price tag on that, 447 00:26:36,280 --> 00:26:39,240 and you can say, OK, how many schools can we build? 448 00:26:39,240 --> 00:26:41,920 And I say, OK, that's your opportunity cost. 449 00:26:41,920 --> 00:26:45,040 You can either let a lot of money. 450 00:26:45,040 --> 00:26:47,240 So we've been doing the math, and it's 451 00:26:47,240 --> 00:26:50,760 hundreds of millions, if not billions of dollars flow 452 00:26:50,760 --> 00:26:53,680 into the hands of who knows who. 453 00:26:53,680 --> 00:26:57,200 Or you could build dozens of new schools. 454 00:26:57,200 --> 00:27:00,040 So the choice is yours. 455 00:27:00,040 --> 00:27:02,120 But now you know my perspective. 456 00:27:02,120 --> 00:27:11,160 But anywho, so those are the sales predictions for 2022. 457 00:27:11,160 --> 00:27:14,840 And what I always tell people here in the Meetup group 458 00:27:14,840 --> 00:27:21,600 is always take numbers on these data points at face value. 459 00:27:21,600 --> 00:27:26,160 So CNBC has put forward their prediction, 31 billion, 460 00:27:26,160 --> 00:27:27,560 4 billion. 461 00:27:27,560 --> 00:27:30,200 But it would be nice to see, OK, what 462 00:27:30,200 --> 00:27:32,400 was their underlying data? 463 00:27:32,400 --> 00:27:35,160 What was their forecasting model? 464 00:27:35,160 --> 00:27:37,120 How did they come to these conclusions? 465 00:27:37,120 --> 00:27:40,080 What assumptions did they make along the way? 466 00:27:40,080 --> 00:27:42,920 And so that's what we do here at the Cannabis Data Science 467 00:27:42,920 --> 00:27:44,400 Meetup group. 468 00:27:44,400 --> 00:27:47,120 We're going to do an open source forecast. 469 00:27:47,120 --> 00:27:49,520 So open source, open data. 470 00:27:49,520 --> 00:27:51,480 So that way, all of you can check it. 471 00:27:51,480 --> 00:27:52,800 We can all improve it. 472 00:27:52,800 --> 00:27:57,880 And then we can apply the whole open source mentality, 473 00:27:57,880 --> 00:28:00,160 where we can improve upon each other. 474 00:28:00,160 --> 00:28:04,200 So what I'm putting forth is not even a good. 475 00:28:04,200 --> 00:28:09,280 It's just it's in a crude initial draft of a forecast. 476 00:28:09,280 --> 00:28:14,760 But I always like to say something, a forecast, 477 00:28:14,760 --> 00:28:19,240 any forecast is better than no forecast. 478 00:28:19,240 --> 00:28:22,560 You can debate that, but that's my opinion. 479 00:28:22,560 --> 00:28:29,960 So without further ado, let's build a forecasting model 480 00:28:29,960 --> 00:28:31,040 here. 481 00:28:31,040 --> 00:28:36,440 And so I always like to give a shout out to John Sylvia 482 00:28:36,440 --> 00:28:38,760 and Azair Iqbal. 483 00:28:38,760 --> 00:28:43,000 They were two professors of mine at UNC Charlotte. 484 00:28:43,000 --> 00:28:46,840 And they wrote a book, Economic and Business Forecasting. 485 00:28:46,840 --> 00:28:49,360 I just keep this with me at all times. 486 00:28:49,360 --> 00:28:53,320 This is a really simple, easy to read book 487 00:28:53,320 --> 00:28:57,360 that is a great foundation for forecasting. 488 00:28:57,360 --> 00:29:00,920 So just would highly recommend this book. 489 00:29:00,920 --> 00:29:05,200 And so long story short, for long term forecasting, 490 00:29:05,200 --> 00:29:08,720 which is 2022, that's a whole year ahead. 491 00:29:08,720 --> 00:29:11,240 So we've done eight theoretical models 492 00:29:11,240 --> 00:29:14,640 for short term forecasts, which are useful for, 493 00:29:14,640 --> 00:29:18,120 which are incredibly good for month ahead, 494 00:29:18,120 --> 00:29:20,480 three month ahead forecasts. 495 00:29:20,480 --> 00:29:24,960 But once you get beyond three months or so, 496 00:29:24,960 --> 00:29:27,000 we're looking at a year here. 497 00:29:27,000 --> 00:29:31,240 And you want a bit more of a structural model 498 00:29:31,240 --> 00:29:35,560 that can capture some economic relationships. 499 00:29:35,560 --> 00:29:41,560 At the same time, they recommend keep it sophisticatedly simple. 500 00:29:41,560 --> 00:29:48,160 So it shouldn't be too technical or complex. 501 00:29:48,160 --> 00:29:52,800 And it shouldn't ignore economic theory. 502 00:29:52,800 --> 00:29:56,240 So you basically want something simple, 503 00:29:56,240 --> 00:29:59,320 but still incorporates economic theory. 504 00:29:59,320 --> 00:30:03,240 So not too complicated, but not too simple 505 00:30:03,240 --> 00:30:05,720 in ignoring economics. 506 00:30:05,720 --> 00:30:09,280 And then reading a good book here, 507 00:30:09,280 --> 00:30:13,760 and John Maynard Keynes is known for a different quote 508 00:30:13,760 --> 00:30:15,520 about the long run. 509 00:30:15,520 --> 00:30:19,800 But I thought this quote was a bit more optimistic. 510 00:30:19,800 --> 00:30:24,400 So in the long run, almost anything is possible. 511 00:30:24,400 --> 00:30:29,880 So I thought this ties in today because we're talking 512 00:30:29,880 --> 00:30:32,040 about states coming online. 513 00:30:32,040 --> 00:30:35,640 So if we're just using an eight theoretical approach, 514 00:30:35,640 --> 00:30:37,920 it's going to be hard to forecast 515 00:30:37,920 --> 00:30:41,120 these states coming online. 516 00:30:41,120 --> 00:30:44,200 So that's just a bit of background. 517 00:30:44,200 --> 00:30:46,800 Here's the actual model. 518 00:30:46,800 --> 00:30:52,320 So essentially, this is super simple, probably too simple, 519 00:30:52,320 --> 00:30:56,960 probably more economic variables can be included. 520 00:30:56,960 --> 00:31:00,440 In their book, Sylvia and Iqbal say 521 00:31:00,440 --> 00:31:04,320 they're using forecasting models with hundreds of variables. 522 00:31:04,320 --> 00:31:08,520 So this has five. 523 00:31:08,520 --> 00:31:12,000 So we could probably add more variables here 524 00:31:12,000 --> 00:31:13,760 to get this more complex. 525 00:31:13,760 --> 00:31:16,000 But you got to start somewhere. 526 00:31:16,000 --> 00:31:18,480 So these were variables that we have. 527 00:31:18,480 --> 00:31:23,000 And we can forecast them into the future pretty well. 528 00:31:23,000 --> 00:31:26,080 So what are we saying here? 529 00:31:26,080 --> 00:31:32,560 I'm basically saying, OK, sales is a function of population. 530 00:31:32,560 --> 00:31:36,480 The more people you have in the state, the higher demand is, 531 00:31:36,480 --> 00:31:40,280 the higher sales are going to be. 532 00:31:40,280 --> 00:31:44,280 Whether or not the state has adult use or not, 533 00:31:44,280 --> 00:31:47,480 just empirically, we've seen that states that 534 00:31:47,480 --> 00:31:52,120 have enacted adult use see their sales increase. 535 00:31:52,120 --> 00:31:54,880 So there must be something going on there, 536 00:31:54,880 --> 00:31:56,440 perhaps on the supply side. 537 00:31:56,440 --> 00:31:59,840 Well, actually, I mean, that's both a supply and a demand 538 00:31:59,840 --> 00:32:02,720 side variable. 539 00:32:02,720 --> 00:32:06,680 Then a couple more things that I was going to include 540 00:32:06,680 --> 00:32:11,640 were the months that this date has permitted medicinal. 541 00:32:11,640 --> 00:32:16,120 And I was thinking this would factor in learning 542 00:32:16,120 --> 00:32:17,400 by doing. 543 00:32:17,400 --> 00:32:20,520 So as you were talking about in New York, 544 00:32:20,520 --> 00:32:23,440 New York's built up the infrastructure 545 00:32:23,440 --> 00:32:28,240 to be able to roll out adult use fairly quickly. 546 00:32:28,240 --> 00:32:32,680 So I was thinking the months that it's had medicinal 547 00:32:32,680 --> 00:32:36,200 may factor into the amount of sales. 548 00:32:36,200 --> 00:32:41,320 Similarly, the months that the state has had adult use. 549 00:32:41,320 --> 00:32:49,240 So for example, Oklahoma and Nevada, 550 00:32:49,240 --> 00:32:53,040 and probably more, were states that enacted adult use, 551 00:32:53,040 --> 00:32:57,120 but the market couldn't quite, the supply 552 00:32:57,120 --> 00:33:00,440 couldn't meet up with the demand right out of the gates. 553 00:33:00,440 --> 00:33:03,640 So cannabis takes some time to grow. 554 00:33:03,640 --> 00:33:09,880 So you need some months to go by for the supply to build up. 555 00:33:09,880 --> 00:33:13,880 Plus, we would expect in the long term, 556 00:33:13,880 --> 00:33:17,120 costs are going to be coming down. 557 00:33:17,120 --> 00:33:20,880 People are going to be inventing new technologies. 558 00:33:20,880 --> 00:33:25,680 So this is sort of an abstract way to cover that. 559 00:33:25,680 --> 00:33:30,160 And then I just tossed in a month dummy variable 560 00:33:30,160 --> 00:33:34,600 just to control for a bit of seasonality. 561 00:33:34,600 --> 00:33:38,800 And not every month is the same. 562 00:33:38,800 --> 00:33:45,520 So without further ado, let's go ahead and estimate 563 00:33:45,520 --> 00:33:48,240 this model here. 564 00:33:48,240 --> 00:33:51,840 So I'm going to just be using Python, 565 00:33:51,840 --> 00:33:54,400 but I always think the data is more 566 00:33:54,400 --> 00:33:56,480 interesting than the software. 567 00:33:56,480 --> 00:34:00,440 So this is just a means to an end. 568 00:34:00,440 --> 00:34:03,880 So for example, if you're going to write this in R, 569 00:34:03,880 --> 00:34:07,240 it should be possible. 570 00:34:07,240 --> 00:34:09,920 So this is just Python. 571 00:34:09,920 --> 00:34:15,040 So just going to read in a handful of useful packages, 572 00:34:15,040 --> 00:34:17,120 just to name a few. 573 00:34:17,120 --> 00:34:21,760 Matplotlib for plotting, Pandas for data wrangling, 574 00:34:21,760 --> 00:34:25,240 stats models for doing statistics, 575 00:34:25,240 --> 00:34:30,560 GrabPopulation from the Federal Reserve, FedFred. 576 00:34:30,560 --> 00:34:33,000 So those are some useful ones, and the rest 577 00:34:33,000 --> 00:34:37,240 are just sort of software manipulation. 578 00:34:37,240 --> 00:34:41,360 Just going to define a couple of handy functions. 579 00:34:41,360 --> 00:34:46,800 So this function will just get data from the Federal Reserve 580 00:34:46,800 --> 00:34:49,400 whenever we need population data. 581 00:34:49,400 --> 00:34:52,120 And then this is just a little helper function, 582 00:34:52,120 --> 00:34:57,640 just to find the months elapsed. 583 00:34:57,640 --> 00:35:00,880 And then just going to read in the API key 584 00:35:00,880 --> 00:35:05,360 and format the plots. 585 00:35:05,360 --> 00:35:10,360 OK, as I said, the data is more interesting than the software. 586 00:35:10,360 --> 00:35:14,560 So let's look at some of this data here. 587 00:35:14,560 --> 00:35:17,600 So in prior weeks, we've just primarily 588 00:35:17,600 --> 00:35:19,560 been working with Excel workbooks, 589 00:35:19,560 --> 00:35:23,440 just because it's pretty simple to create an Excel workbook 590 00:35:23,440 --> 00:35:24,920 and look at it. 591 00:35:24,920 --> 00:35:29,920 But all these software developers and data scientists 592 00:35:29,920 --> 00:35:31,320 probably are already familiar. 593 00:35:31,320 --> 00:35:37,520 This is JSON, so JavaScript Object Notation. 594 00:35:37,520 --> 00:35:40,680 And it's just another way of displaying data. 595 00:35:40,680 --> 00:35:47,720 And it's quite handy, because you can do nested data. 596 00:35:47,720 --> 00:35:50,640 So I love JSON. 597 00:35:50,640 --> 00:35:57,400 So I think it really helps to think about these objects. 598 00:35:57,400 --> 00:36:03,480 So I'd like to use the analogy of a duck or ducks. 599 00:36:03,480 --> 00:36:06,600 So these are all states. 600 00:36:06,600 --> 00:36:10,120 So say you've got a group of ducklings. 601 00:36:10,120 --> 00:36:13,280 You've got a whole bunch of ducklings. 602 00:36:13,280 --> 00:36:17,120 But each duck, each duckling, is going 603 00:36:17,120 --> 00:36:19,440 to have different properties. 604 00:36:19,440 --> 00:36:22,680 So these are all our states here. 605 00:36:22,680 --> 00:36:26,440 And we can keep track of different properties 606 00:36:26,440 --> 00:36:29,080 for all the different states. 607 00:36:29,080 --> 00:36:33,680 And what I've done is basically just indicated, OK, 608 00:36:33,680 --> 00:36:39,960 are the states permitting medicinal or adult use? 609 00:36:39,960 --> 00:36:43,760 And then when was medicinal permitted? 610 00:36:43,760 --> 00:36:46,560 And when was adult use permitted? 611 00:36:46,560 --> 00:36:50,640 And this is easier said than done. 612 00:36:50,640 --> 00:36:54,080 So this is where I say this can be improved upon. 613 00:36:54,080 --> 00:36:56,920 So I would actually highly recommend you do this, 614 00:36:56,920 --> 00:36:58,600 because I did this really quickly. 615 00:36:58,600 --> 00:37:03,520 And I still need to go back over and double check everything. 616 00:37:03,520 --> 00:37:06,960 But check out the source, because so for example, 617 00:37:06,960 --> 00:37:13,840 this just says, oh, California permitted medicinal in 1996. 618 00:37:13,840 --> 00:37:17,600 Well, we're going to have to hit the history books and newspapers 619 00:37:17,600 --> 00:37:22,080 and find out what month in 1996. 620 00:37:22,080 --> 00:37:25,720 And when the actual stores were opened. 621 00:37:25,720 --> 00:37:32,520 So I'm trying to set these dates when the actual stores were 622 00:37:32,520 --> 00:37:33,320 opened. 623 00:37:33,320 --> 00:37:36,920 So for example, our representative here 624 00:37:36,920 --> 00:37:42,400 from Colorado, I'm not certain I got these dates correct. 625 00:37:42,400 --> 00:37:45,120 I tried to do my best of identifying 626 00:37:45,120 --> 00:37:50,160 when dispensaries actually opened their doors. 627 00:37:50,160 --> 00:37:56,520 So if you notice any mistakes here, please correct. 628 00:37:56,520 --> 00:37:59,360 Because we've pointed out in the past 629 00:37:59,360 --> 00:38:03,880 that measurement error can lead to bias in your results. 630 00:38:03,880 --> 00:38:07,960 So if, say, some of these, I'm using the date 631 00:38:07,960 --> 00:38:10,360 where the legislation was passed, 632 00:38:10,360 --> 00:38:13,400 but not actually when the dispensaries were opened, 633 00:38:13,400 --> 00:38:15,440 that's measurement error and will 634 00:38:15,440 --> 00:38:17,600 lead to bias in our results. 635 00:38:17,600 --> 00:38:24,000 So this is our first assumption along the way, major assumption, 636 00:38:24,000 --> 00:38:29,440 is that sales start when things are permitted. 637 00:38:29,440 --> 00:38:35,920 But you'll see, if you start reading through these sources, 638 00:38:35,920 --> 00:38:43,280 you'll see that various states rolled things out quickly. 639 00:38:43,280 --> 00:38:45,680 And sometimes things were delayed. 640 00:38:45,680 --> 00:38:48,920 So maybe they permitted cannabis at a certain point, 641 00:38:48,920 --> 00:38:54,080 but the dispensaries didn't open until years later. 642 00:38:54,080 --> 00:38:56,720 So long story short, I've done my best 643 00:38:56,720 --> 00:39:00,640 to identify when each state permitted. 644 00:39:04,400 --> 00:39:07,720 And then the other data set we'll be using 645 00:39:07,720 --> 00:39:11,760 is this is basically all the sales data 646 00:39:11,760 --> 00:39:14,720 that we've collected in 2021. 647 00:39:14,720 --> 00:39:20,920 So we were collecting Colorado sales, Illinois sales, 648 00:39:20,920 --> 00:39:24,160 Massachusetts sales. 649 00:39:24,160 --> 00:39:27,880 Notice Massachusetts was closed for a month there. 650 00:39:27,880 --> 00:39:33,120 But anywho, there are still more data points to be collected. 651 00:39:33,120 --> 00:39:37,560 But these were the ones that we diligently collected. 652 00:39:37,560 --> 00:39:44,080 So as a good Bayesian, we can use our prior knowledge 653 00:39:44,080 --> 00:39:47,640 or prior data in our forecasts here. 654 00:39:47,640 --> 00:39:50,360 So basically, we'll be using the combination 655 00:39:50,360 --> 00:39:55,080 of historic sales data to train our model. 656 00:39:55,080 --> 00:39:58,840 And then we'll use these variables 657 00:39:58,840 --> 00:40:01,600 to try to predict for states where 658 00:40:01,600 --> 00:40:06,080 we don't have good historic data. 659 00:40:06,080 --> 00:40:09,400 What's the data source of the sales data? 660 00:40:09,400 --> 00:40:12,800 Yes, I still need to finish adding them. 661 00:40:12,800 --> 00:40:16,600 I've started to add some here on the sources tab. 662 00:40:16,600 --> 00:40:19,160 There's still some missing. 663 00:40:19,160 --> 00:40:21,800 They're scattered throughout our 2021. 664 00:40:21,800 --> 00:40:24,080 What I'll do is I've ended today. 665 00:40:24,080 --> 00:40:26,200 I'll update all these sources here. 666 00:40:26,200 --> 00:40:31,320 So that way, you can have the sources. 667 00:40:31,320 --> 00:40:35,560 But essentially, they're all public data sources 668 00:40:35,560 --> 00:40:41,000 where those various states are publishing various data points. 669 00:40:41,000 --> 00:40:43,920 Great, thanks. 670 00:40:43,920 --> 00:40:45,240 But good question. 671 00:40:45,240 --> 00:40:47,520 It's always important to show your sources. 672 00:40:47,520 --> 00:40:53,960 And that's what we're doing here, open data, open source. 673 00:40:53,960 --> 00:40:54,880 Awesome. 674 00:40:54,880 --> 00:40:58,880 So we've got our data. 675 00:40:58,880 --> 00:41:00,800 Let's go ahead and read it in. 676 00:41:04,200 --> 00:41:06,120 Awesome. 677 00:41:06,120 --> 00:41:09,040 Just take a quick look at it. 678 00:41:09,040 --> 00:41:09,640 There it is. 679 00:41:09,640 --> 00:41:11,880 There's the exact same data we were just looking at. 680 00:41:14,800 --> 00:41:20,000 Now, time to start doing data science. 681 00:41:20,000 --> 00:41:22,400 So one of my favorite things to do 682 00:41:22,400 --> 00:41:27,040 is create and retrieve supplementary variables, 683 00:41:27,040 --> 00:41:31,200 because you'd be surprised what you can calculate. 684 00:41:31,200 --> 00:41:37,400 So for example, we're going to basically use 685 00:41:37,400 --> 00:41:42,520 when the state permitted medicinal or adult use. 686 00:41:42,520 --> 00:41:46,840 That way, we can calculate how many months the states 687 00:41:46,840 --> 00:41:50,960 had medicinal or adult use. 688 00:41:50,960 --> 00:41:55,680 So don't want to spend too much time on the code. 689 00:41:55,680 --> 00:41:59,920 But basically, just checking, OK, 690 00:41:59,920 --> 00:42:02,680 is the state an adult use state? 691 00:42:02,680 --> 00:42:10,040 And then I'm calculating down here how many months elapsed 692 00:42:10,040 --> 00:42:13,280 since then for each point in time. 693 00:42:13,280 --> 00:42:19,320 And then just creating a set of month effects. 694 00:42:19,320 --> 00:42:25,520 So we get to look at an observation. 695 00:42:25,520 --> 00:42:28,960 So for example, here is Washington. 696 00:42:28,960 --> 00:42:36,920 The last data point we have for there is October of 2021. 697 00:42:36,920 --> 00:42:44,160 And we say that, OK, there's been 160 months of adult use 698 00:42:44,160 --> 00:42:47,760 and 285 months of medicinal use. 699 00:42:47,760 --> 00:42:54,560 And so we're hypothesizing that that has been the case 700 00:42:54,560 --> 00:43:00,880 and that that has an effect on total sales. 701 00:43:00,880 --> 00:43:05,640 So those are our main metrics here. 702 00:43:05,640 --> 00:43:12,280 And then month effects are basically just 0 or 1 703 00:43:12,280 --> 00:43:15,320 if it is a certain month. 704 00:43:15,320 --> 00:43:19,040 And you need a baseline comparison month. 705 00:43:19,040 --> 00:43:21,920 So I've excluded January. 706 00:43:21,920 --> 00:43:24,920 So basically, our month effects will tell us 707 00:43:24,920 --> 00:43:30,920 how different a given month is from January. 708 00:43:30,920 --> 00:43:37,920 So here, the parameter beta 5, it's actually 709 00:43:37,920 --> 00:43:46,640 beta 5 through 15 because we have to include 11 dummy 710 00:43:46,640 --> 00:43:48,400 variables. 711 00:43:48,400 --> 00:43:52,240 So if we're going to do a much coefficient, we'll say, OK, 712 00:43:52,240 --> 00:43:55,880 this is how much different February is from January, 713 00:43:55,880 --> 00:43:59,240 how much different July is from January. 714 00:43:59,240 --> 00:44:03,520 So the idea is you should be able to exclude 715 00:44:03,520 --> 00:44:08,400 any given month and our estimates 716 00:44:08,400 --> 00:44:10,760 for the other parameters should be the same. 717 00:44:14,680 --> 00:44:15,400 Should be. 718 00:44:15,400 --> 00:44:17,920 I have to double check my statistics here. 719 00:44:17,920 --> 00:44:21,920 Without further ado, let's estimate this regression model. 720 00:44:21,920 --> 00:44:23,600 So here it is. 721 00:44:23,600 --> 00:44:27,200 So it's basically just a constant, a coefficient 722 00:44:27,200 --> 00:44:32,720 on population, coefficient on a dummy variable, 0 or 1 723 00:44:32,720 --> 00:44:36,600 if the state has adult use, the months 724 00:44:36,600 --> 00:44:39,520 that it's been medicinal, the months 725 00:44:39,520 --> 00:44:42,120 that it's been adult use. 726 00:44:42,120 --> 00:44:46,200 And this is called an interaction with adult use. 727 00:44:46,200 --> 00:44:50,480 So basically, if the state doesn't have adult use, 728 00:44:50,480 --> 00:44:54,120 then this will just take a value of 0. 729 00:44:54,120 --> 00:44:57,200 I thought that was reasonable. 730 00:44:57,200 --> 00:45:01,800 And then the dummy variable for the month. 731 00:45:01,800 --> 00:45:09,720 So here is the regression of sales on these variables. 732 00:45:09,720 --> 00:45:14,480 And I just did this cursory earlier, 733 00:45:14,480 --> 00:45:18,040 so I actually haven't analyzed this regression. 734 00:45:18,040 --> 00:45:21,160 So just going to do it right here for you. 735 00:45:21,160 --> 00:45:25,560 So first things first, I'd like to look at the r squared. 736 00:45:25,560 --> 00:45:29,840 You can roughly interpret this as how much variation 737 00:45:29,840 --> 00:45:33,640 are we explaining with our variables. 738 00:45:33,640 --> 00:45:38,440 And so you would love an r squared close to 1 739 00:45:38,440 --> 00:45:42,840 because you're explaining almost all of the variation 740 00:45:42,840 --> 00:45:44,440 at that point. 741 00:45:44,440 --> 00:45:53,200 0.47 is actually not the worst for a structural regression. 742 00:45:53,200 --> 00:45:56,640 I've seen r squared is much lower, 743 00:45:56,640 --> 00:45:59,440 but it would also be nice to have an r squared much higher. 744 00:45:59,440 --> 00:46:05,880 So what you can do is you can start to add variables. 745 00:46:05,880 --> 00:46:10,040 And what you can do is you can actually, 746 00:46:10,040 --> 00:46:13,800 there are some statistical tests you can do to see, 747 00:46:13,800 --> 00:46:18,000 you can see, OK, let's try to add some variables 748 00:46:18,000 --> 00:46:23,800 and basically penalize ourselves for adding more variables 749 00:46:23,800 --> 00:46:26,520 while rewarding ourselves if we're 750 00:46:26,520 --> 00:46:29,320 able to explain more variation. 751 00:46:29,320 --> 00:46:33,800 So long story short, you can compare models. 752 00:46:33,800 --> 00:46:36,560 And I would highly recommend all of you to, 753 00:46:36,560 --> 00:46:44,600 maybe you can think of creative independent variables here 754 00:46:44,600 --> 00:46:46,400 that you would like to add. 755 00:46:46,400 --> 00:46:49,360 So personally, I would just start 756 00:46:49,360 --> 00:46:53,880 thinking of demand and supply side variables 757 00:46:53,880 --> 00:46:57,000 and see if we can't include those. 758 00:46:57,000 --> 00:47:01,120 Keep in mind, if you add a variable, 759 00:47:01,120 --> 00:47:06,920 you're going to have to forecast it into 2022. 760 00:47:06,920 --> 00:47:12,440 For population, I used the most naive forecast you can have. 761 00:47:12,440 --> 00:47:16,640 And I just said, OK, the population in 2022 762 00:47:16,640 --> 00:47:21,040 is going to be the same as it's going to be in 2021. 763 00:47:21,040 --> 00:47:25,360 Another big assumption, you can improve upon this model 764 00:47:25,360 --> 00:47:28,720 by forecasting population. 765 00:47:28,720 --> 00:47:33,200 So if you've, once again, it's going 766 00:47:33,200 --> 00:47:34,520 to add a bit more uncertainty. 767 00:47:34,520 --> 00:47:38,600 But it may result in a better forecast 768 00:47:38,600 --> 00:47:43,320 because it's a stretch to assume that population is going 769 00:47:43,320 --> 00:47:45,240 to remain constant in every single state 770 00:47:45,240 --> 00:47:46,440 across the country. 771 00:47:46,440 --> 00:47:47,680 That's not realistic. 772 00:47:47,680 --> 00:47:53,200 So I like to kind of preface the work I'm doing here 773 00:47:53,200 --> 00:47:54,400 in this meetup. 774 00:47:54,400 --> 00:47:56,920 It's a lot of a demonstration. 775 00:47:56,920 --> 00:48:01,480 So please take these numbers at face value 776 00:48:01,480 --> 00:48:05,240 and know that all of these forecasts can and should 777 00:48:05,240 --> 00:48:06,960 be improved upon. 778 00:48:06,960 --> 00:48:10,040 So for example, if you're actually 779 00:48:10,040 --> 00:48:12,360 doing something for the state of Colorado, 780 00:48:12,360 --> 00:48:17,760 definitely want a bit more rigor than I'm using today. 781 00:48:17,760 --> 00:48:24,280 But it's more just a thought exercise, a proof of concept 782 00:48:24,280 --> 00:48:25,840 that we're doing today. 783 00:48:25,840 --> 00:48:28,520 So just for the sake of time, just 784 00:48:28,520 --> 00:48:32,600 going to use these five variables plus the month 785 00:48:32,600 --> 00:48:40,480 effects and know that there's ways to improve upon this. 786 00:48:40,480 --> 00:48:46,560 So let's see if we can't do any rough interpretations 787 00:48:46,560 --> 00:48:48,560 of the data here. 788 00:48:48,560 --> 00:48:53,480 So for example, our coefficient on population 789 00:48:53,480 --> 00:49:00,320 is 6.6, so roughly speaking, for each additional person 790 00:49:00,320 --> 00:49:09,960 in a state, you would expect around $6.60 additional sales. 791 00:49:09,960 --> 00:49:14,800 I've seen a lot of people use adult population, right? 792 00:49:14,800 --> 00:49:18,480 Because just say somebody's born, 793 00:49:18,480 --> 00:49:22,760 they're not going to be consuming cannabis in 2022. 794 00:49:22,760 --> 00:49:28,360 So long story short, you could improve upon this 795 00:49:28,360 --> 00:49:33,160 by perhaps using a measure for adult population. 796 00:49:33,160 --> 00:49:38,960 Maybe I missed something, but where did you factor in price? 797 00:49:38,960 --> 00:49:41,320 I thought we were talking numbers. 798 00:49:41,320 --> 00:49:43,320 Excellent, excellent question. 799 00:49:43,320 --> 00:49:45,360 I'm not factoring in price. 800 00:49:45,360 --> 00:49:51,800 That could be and is an omitted variable. 801 00:49:51,800 --> 00:49:57,920 So as we've seen in states such as there's 802 00:49:57,920 --> 00:50:05,960 a large price discrepancy between prices in, say, 803 00:50:05,960 --> 00:50:11,240 the two states we have price data are Oregon 804 00:50:11,240 --> 00:50:15,560 and I want to say Massachusetts perhaps, or yes, in Maine. 805 00:50:15,560 --> 00:50:18,320 So we have prices in a handful of states 806 00:50:18,320 --> 00:50:22,680 and we could get prices in Washington if we wanted. 807 00:50:22,680 --> 00:50:24,440 Long story short is we don't really 808 00:50:24,440 --> 00:50:30,000 have price data for most of these states. 809 00:50:30,000 --> 00:50:33,920 So then I'm not clear what the population coefficient means. 810 00:50:33,920 --> 00:50:38,440 Would it be number of sales rather than dollars? 811 00:50:38,440 --> 00:50:39,520 It would just be dollars. 812 00:50:39,520 --> 00:50:46,320 So basically, our dependent variable here is sales. 813 00:50:46,320 --> 00:50:49,400 So this is just sales in dollars. 814 00:50:49,400 --> 00:50:56,720 So the total gross amount grossed by cannabis sales 815 00:50:56,720 --> 00:50:59,640 in a given state. 816 00:50:59,640 --> 00:51:01,000 Does that answer your question? 817 00:51:04,840 --> 00:51:06,640 So basically, what I'm saying here 818 00:51:06,640 --> 00:51:11,200 is for each additional person in the state, 819 00:51:11,200 --> 00:51:14,360 you would expect the total sales in the state 820 00:51:14,360 --> 00:51:18,880 to increase by $6.60. 821 00:51:18,880 --> 00:51:22,480 How much actual cannabis that is, like you said, 822 00:51:22,480 --> 00:51:24,520 depends on the price. 823 00:51:24,520 --> 00:51:31,720 So for example, $6 goes a lot further in Oregon 824 00:51:31,720 --> 00:51:35,840 than it does in Massachusetts or Maine 825 00:51:35,840 --> 00:51:40,120 because the prices in Oregon are like a third of what they 826 00:51:40,120 --> 00:51:42,680 are in Massachusetts and Maine. 827 00:51:42,680 --> 00:51:47,440 So long story short, I would love 828 00:51:47,440 --> 00:51:51,120 to include a measure of price for each state, 829 00:51:51,120 --> 00:51:55,040 like for example, maybe average price in a given state. 830 00:51:55,040 --> 00:51:58,440 I just don't have that data. 831 00:51:58,440 --> 00:52:03,200 And I would have to forecast that data in the 2022. 832 00:52:03,200 --> 00:52:08,000 So once again, there are companies 833 00:52:08,000 --> 00:52:10,320 out there that actually do a really good job 834 00:52:10,320 --> 00:52:13,320 at measuring price. 835 00:52:13,320 --> 00:52:15,600 A lot of that's private data. 836 00:52:15,600 --> 00:52:18,880 However, they could improve upon these forecasts 837 00:52:18,880 --> 00:52:22,240 by including a metric for price. 838 00:52:25,560 --> 00:52:26,600 Excellent point. 839 00:52:29,080 --> 00:52:33,880 Surprisingly, we have this negative coefficient 840 00:52:33,880 --> 00:52:36,480 on adult use. 841 00:52:36,480 --> 00:52:39,400 So that would say, oh, having adult use actually 842 00:52:39,400 --> 00:52:41,440 lowers your baseline. 843 00:52:41,440 --> 00:52:44,040 But we've got a positive coefficient 844 00:52:44,040 --> 00:52:46,880 on the months of adult use. 845 00:52:46,880 --> 00:52:49,880 So it seems that the more and more months 846 00:52:49,880 --> 00:52:53,360 that you have adult use, the higher and higher 847 00:52:53,360 --> 00:52:55,600 your sales are going to be. 848 00:52:55,600 --> 00:53:00,360 Similar for medicinal, however, the coefficient 849 00:53:00,360 --> 00:53:05,520 is smaller than adult use, around a third of the size. 850 00:53:05,520 --> 00:53:11,160 So the effect of having adult use is almost three times 851 00:53:11,160 --> 00:53:13,000 as great per month as medicinal. 852 00:53:17,560 --> 00:53:25,200 If you're a frequentist, they're looking fairly significant here. 853 00:53:25,200 --> 00:53:30,760 One thing that you'll see is not many of these month effects 854 00:53:30,760 --> 00:53:38,520 are statistically significant, except for perhaps July 855 00:53:38,520 --> 00:53:43,560 and August at the 95% confidence level. 856 00:53:43,560 --> 00:53:46,400 So basically, you really couldn't 857 00:53:46,400 --> 00:53:50,840 conclude that any month is different in sales. 858 00:53:50,840 --> 00:53:54,720 Keep in mind, our training data is limited. 859 00:53:54,720 --> 00:53:56,480 I showed you that earlier. 860 00:53:56,480 --> 00:54:04,600 Our training data doesn't include all of the states. 861 00:54:04,600 --> 00:54:08,640 So I need to expand our training data, 862 00:54:08,640 --> 00:54:14,800 because there is still public data that's not been collected. 863 00:54:14,800 --> 00:54:18,320 So that's another way to expand on our forecasts, 864 00:54:18,320 --> 00:54:24,240 is we can add more training data to just train a better 865 00:54:24,240 --> 00:54:26,680 and better model here. 866 00:54:26,680 --> 00:54:31,160 So long story short, I'm including them anyways. 867 00:54:31,160 --> 00:54:34,880 But you may want to try to re-estimate this model 868 00:54:34,880 --> 00:54:38,400 without month effects and see maybe 869 00:54:38,400 --> 00:54:41,240 you don't need these month effects. 870 00:54:41,240 --> 00:54:43,280 I'm including them because they're 871 00:54:43,280 --> 00:54:47,080 one thing that we can forecast with 100% certainty. 872 00:54:47,080 --> 00:54:50,720 We always know that in October of 2022, 873 00:54:50,720 --> 00:54:52,480 it's going to be October. 874 00:54:52,480 --> 00:54:55,040 So we know that with 100% certainty. 875 00:54:55,040 --> 00:55:03,280 So I think they're a nice go-to when forecasting. 876 00:55:03,280 --> 00:55:03,960 Awesome. 877 00:55:03,960 --> 00:55:06,400 So without further ado, just going 878 00:55:06,400 --> 00:55:09,600 to define the parameters here. 879 00:55:09,600 --> 00:55:13,600 And here's just a whole bunch of software. 880 00:55:13,600 --> 00:55:19,840 But basically, what I'm doing here is, OK, 881 00:55:19,840 --> 00:55:24,680 we've got this list of states here. 882 00:55:24,680 --> 00:55:31,440 So I'm just going to iterate over all of the states, 883 00:55:31,440 --> 00:55:35,200 find out when adult use was permitted in each state, 884 00:55:35,200 --> 00:55:38,120 when medicinal was permitted in each state, 885 00:55:38,120 --> 00:55:42,280 get the population of 2021, which was actually just 886 00:55:42,280 --> 00:55:44,640 published a few weeks ago. 887 00:55:44,640 --> 00:55:49,680 So long story short, we've got those variables for each state. 888 00:55:49,680 --> 00:55:56,280 And then I'm just going to iterate over our forecast 889 00:55:56,280 --> 00:56:00,280 horizon, which is 2022. 890 00:56:00,280 --> 00:56:04,840 So iterating over each state, iterating over each date, 891 00:56:04,840 --> 00:56:08,360 and then just a handful of if statements 892 00:56:08,360 --> 00:56:14,280 to figure out, OK, what are the variables we need to include. 893 00:56:14,280 --> 00:56:19,920 But nothing too fancy, really just 894 00:56:19,920 --> 00:56:23,000 this is where the magic happens, where we actually 895 00:56:23,000 --> 00:56:27,720 estimate the forecast, beta naught plus beta 1 times 896 00:56:27,720 --> 00:56:31,080 population, beta 2 times adult use, 897 00:56:31,080 --> 00:56:35,000 beta 3 medicinal, so on and so forth, 898 00:56:35,000 --> 00:56:39,320 just as we specified in our regression model. 899 00:56:39,320 --> 00:56:42,160 So let's do that. 900 00:56:42,160 --> 00:56:45,400 So basically, I'm just plugging in variables 901 00:56:45,400 --> 00:56:53,560 into the regression model that we just estimated. 902 00:56:53,560 --> 00:56:57,160 Takes it a handful of seconds because I didn't write 903 00:56:57,160 --> 00:57:00,240 the most optimal routine here. 904 00:57:00,240 --> 00:57:03,200 There's a lot of ways you can optimize this. 905 00:57:03,200 --> 00:57:08,160 But as Edward Tufts says, show the data. 906 00:57:08,160 --> 00:57:13,000 So here are, whoops, I didn't quite do that right. 907 00:57:18,120 --> 00:57:19,080 Give me one second here. 908 00:57:19,080 --> 00:57:20,160 I can fix this. 909 00:57:20,160 --> 00:57:31,760 Actually, I think all we need to do is add a time index to this. 910 00:57:43,880 --> 00:57:44,360 Hold on. 911 00:57:44,360 --> 00:57:45,840 Yes, question? 912 00:57:49,040 --> 00:57:50,080 Is there a question at hand? 913 00:57:52,800 --> 00:58:03,440 OK, so we may have to just save this part because we can still 914 00:58:03,440 --> 00:58:05,640 look at, I'm going to try this one other thing 915 00:58:05,640 --> 00:58:08,600 and then we'll just move on. 916 00:58:08,600 --> 00:58:09,120 Awesome. 917 00:58:09,120 --> 00:58:13,400 Just had to add the index just to kind of line everything up. 918 00:58:13,400 --> 00:58:17,880 So this isn't actually the most informative chart ever. 919 00:58:17,880 --> 00:58:20,720 Was there a question? 920 00:58:20,720 --> 00:58:22,240 Some people have to jump. 921 00:58:22,240 --> 00:58:24,080 And I always try to wrap it up. 922 00:58:24,080 --> 00:58:25,840 So I'll be wrapping this up shortly. 923 00:58:25,840 --> 00:58:28,240 And then if you do have to draw, you 924 00:58:28,240 --> 00:58:32,080 can check out the recording on YouTube. 925 00:58:32,080 --> 00:58:35,880 So I'm going to try to add a time index. 926 00:58:35,880 --> 00:58:39,040 And then I'm going to try to do some more 927 00:58:39,040 --> 00:58:42,040 of the recording on YouTube. 928 00:58:42,040 --> 00:58:47,440 So not the best plot in the world because there's 929 00:58:47,440 --> 00:58:49,880 just so many states. 930 00:58:49,880 --> 00:58:55,120 And I think I'm reusing some colors over and over again. 931 00:58:55,120 --> 00:59:00,720 But we've got our forecast for California on top. 932 00:59:00,720 --> 00:59:06,920 Oh, yes, let's just list these because I think the list is 933 00:59:06,920 --> 00:59:10,680 a bit more informative than this big plot. 934 00:59:10,680 --> 00:59:19,400 So here's just the predicted sales by state. 935 00:59:19,400 --> 00:59:25,480 And I'm going to order this by the most sales. 936 00:59:25,480 --> 00:59:29,360 So that way, it's a little easier to read here. 937 00:59:29,360 --> 00:59:32,760 And as you'll notice, our forecasting model 938 00:59:32,760 --> 00:59:35,560 has some wonky predictions. 939 00:59:35,560 --> 00:59:43,160 So basically, Texas does technically have medicinal use. 940 00:59:43,160 --> 00:59:46,280 And they've got a giant population. 941 00:59:46,280 --> 00:59:50,160 So because of the way our forecasting model was 942 00:59:50,160 --> 00:59:56,080 constructed, Texas is taking the number two sales. 943 00:59:56,080 --> 01:00:00,240 I don't think this is realistic per se. 944 01:00:00,240 --> 01:00:03,840 So I think this can be improved upon. 945 01:00:03,840 --> 01:00:06,560 So for example, Florida is number three. 946 01:00:06,560 --> 01:00:10,000 Once again, I'm not certain about this. 947 01:00:10,000 --> 01:00:13,040 But we don't really have good data from Florida. 948 01:00:13,040 --> 01:00:17,800 So if you look at our regression model, all we're really using 949 01:00:17,800 --> 01:00:22,400 is their population, whether they're allowing adult use, 950 01:00:22,400 --> 01:00:25,160 whether they're allowing medicinal use. 951 01:00:25,160 --> 01:00:30,960 So our model can probably be improved upon. 952 01:00:30,960 --> 01:00:33,240 But really, the population is what's 953 01:00:33,240 --> 01:00:38,240 carrying our estimations here. 954 01:00:38,240 --> 01:00:41,560 But like I said, any forecast, I think, 955 01:00:41,560 --> 01:00:43,800 is better than no forecast. 956 01:00:43,800 --> 01:00:50,320 So we're way underestimating Colorado 957 01:00:50,320 --> 01:00:54,520 because we have done eight theoretical forecasts. 958 01:00:54,520 --> 01:00:58,040 And I think our eight theoretical forecast 959 01:00:58,040 --> 01:01:06,640 was that Colorado would be around 1.6 billion plus in 2022. 960 01:01:06,640 --> 01:01:13,000 So this model has some huge shortcomings. 961 01:01:13,000 --> 01:01:19,240 But we just needed to sort of do something 962 01:01:19,240 --> 01:01:24,440 because we just wanted to get a number for each state here. 963 01:01:24,440 --> 01:01:27,360 Because now we can compare our estimates 964 01:01:27,360 --> 01:01:31,520 to CNBC's estimates. 965 01:01:31,520 --> 01:01:37,360 So given our crude forecasting model, 966 01:01:37,360 --> 01:01:45,800 we're predicting 1.5 billion in New York in 2022. 967 01:01:45,800 --> 01:01:49,640 And if you look at my assumptions here, 968 01:01:49,640 --> 01:01:56,040 that is under the assumption that New York permits 969 01:01:56,040 --> 01:02:01,160 adult use on October 1. 970 01:02:01,160 --> 01:02:04,320 My source was this website, which 971 01:02:04,320 --> 01:02:09,760 suggested that New York would do it near the end of 2022. 972 01:02:09,760 --> 01:02:12,880 So that's an assumption built into the model 973 01:02:12,880 --> 01:02:16,840 that can be improved upon. 974 01:02:16,840 --> 01:02:22,000 But long story short, our total estimate for the US, 975 01:02:22,000 --> 01:02:29,480 given all our imperfections along the way, is 26.9 billion. 976 01:02:29,480 --> 01:02:34,160 So now we actually have a number to put on it. 977 01:02:34,160 --> 01:02:41,760 And I would like to say that now we can actually say, OK, 978 01:02:41,760 --> 01:02:44,240 whose model predicts better? 979 01:02:44,240 --> 01:02:46,040 Because we can actually tell. 980 01:02:46,040 --> 01:02:49,600 So we can wait till the end of 2022. 981 01:02:49,600 --> 01:02:56,760 And we can basically calculate the root mean squared error. 982 01:02:56,760 --> 01:03:01,080 And we can basically see whose model predicts better. 983 01:03:01,080 --> 01:03:04,520 Does CNBC's model predict better? 984 01:03:04,520 --> 01:03:08,680 Does this crude model predict better? 985 01:03:08,680 --> 01:03:12,600 Or perhaps you. 986 01:03:12,600 --> 01:03:14,880 You can create your own forecast. 987 01:03:14,880 --> 01:03:16,800 And so I was thinking about maybe we 988 01:03:16,800 --> 01:03:18,640 could do some sort of competition, 989 01:03:18,640 --> 01:03:23,920 where we all could make our own forecasting model. 990 01:03:23,920 --> 01:03:25,920 Maybe we could do a mix. 991 01:03:25,920 --> 01:03:29,920 So I think that may be the best sort of mixed model, where 992 01:03:29,920 --> 01:03:36,000 we use the a theoretical model when we do have sales data, 993 01:03:36,000 --> 01:03:38,800 such as in Colorado. 994 01:03:38,800 --> 01:03:41,720 And then we can use the a theoretical. 995 01:03:41,720 --> 01:03:43,640 I mean, we can use the structural model, 996 01:03:43,640 --> 01:03:48,080 the theoretical model, for states where we don't have good data. 997 01:03:50,720 --> 01:03:55,840 A lot of the states, we don't have fantastic data. 998 01:03:55,840 --> 01:03:57,320 So we still need to look a bit more. 999 01:03:57,320 --> 01:04:01,920 But long story short is a lot of this 1000 01:04:01,920 --> 01:04:05,160 is chalked up to medicinal sales, 1001 01:04:05,160 --> 01:04:09,960 which I'm skeptical will be as high as our estimates are. 1002 01:04:09,960 --> 01:04:14,520 But any estimate's better than no estimate. 1003 01:04:14,520 --> 01:04:18,280 So there we have the numbers. 1004 01:04:18,280 --> 01:04:20,280 We've run a little long. 1005 01:04:20,280 --> 01:04:23,720 So I'm going to go ahead and end the presentation. 1006 01:04:23,720 --> 01:04:26,600 But does anybody have any thoughts or comments? 1007 01:04:26,600 --> 01:04:28,200 Yeah, I'm a little overwhelmed. 1008 01:04:28,200 --> 01:04:31,280 I mean, it's a lot. 1009 01:04:31,280 --> 01:04:38,360 But my first thought is why concatenate medicinal 1010 01:04:38,360 --> 01:04:44,200 and recreational use just to get a total number 1011 01:04:44,200 --> 01:04:50,120 or because they're kind of different. 1012 01:04:50,120 --> 01:04:57,000 Wouldn't a more granular focus? 1013 01:04:57,000 --> 01:04:59,320 I mean, yeah, you want to have an overall number, I guess, 1014 01:04:59,320 --> 01:05:02,360 the farmers, because they're going to be producing it. 1015 01:05:02,360 --> 01:05:07,200 But wouldn't a more granular take on it help? 1016 01:05:07,200 --> 01:05:09,200 Yes, 100%. 1017 01:05:09,200 --> 01:05:11,680 And that sort of brings us back to sort of this day to day 1018 01:05:11,680 --> 01:05:14,480 luge, where I think you're 100% right. 1019 01:05:14,480 --> 01:05:15,480 I think you're right. 1020 01:05:15,480 --> 01:05:16,480 I think you're right. 1021 01:05:16,480 --> 01:05:20,440 I think it would be good to have an estimate of adult use 1022 01:05:20,440 --> 01:05:23,800 in medicinal because I think they're different things. 1023 01:05:23,800 --> 01:05:28,600 I was primarily doing this just to compare to CNBC's number 1024 01:05:28,600 --> 01:05:34,920 because they said that was their estimate of total legal sales. 1025 01:05:34,920 --> 01:05:38,920 So medicinal sales, they're still legal. 1026 01:05:38,920 --> 01:05:42,440 So I was including them. 1027 01:05:42,440 --> 01:05:44,920 And it's really important to have 1028 01:05:44,920 --> 01:05:49,600 a set of estimates for states with medicinal use. 1029 01:05:49,600 --> 01:05:53,680 And it's really tough to distinguish these, right? 1030 01:05:53,680 --> 01:05:57,680 Because, for example, Oklahoma is a medicinal state, 1031 01:05:57,680 --> 01:06:03,160 but it's fairly permissible, whereas other states may only 1032 01:06:03,160 --> 01:06:10,280 permit, they may have heavily restricted medicinal programs. 1033 01:06:10,280 --> 01:06:14,880 So that's why I was saying a lot of these estimates 1034 01:06:14,880 --> 01:06:17,520 because I don't know how much faith I would put in those 1035 01:06:17,520 --> 01:06:22,320 forecasts because I think there's 1036 01:06:22,320 --> 01:06:30,400 a big discrepancies in how the medicinal programs behave 1037 01:06:30,400 --> 01:06:31,800 from state to state. 1038 01:06:31,800 --> 01:06:36,400 So if there's a way to capture that in a variable, 1039 01:06:36,400 --> 01:06:39,800 then we should all brainstorm and we could probably 1040 01:06:39,800 --> 01:06:40,600 improve upon them all. 1041 01:06:40,600 --> 01:06:42,800 But I'll just leave you with this. 1042 01:06:42,800 --> 01:06:44,480 This is a starting point. 1043 01:06:44,480 --> 01:06:45,960 It's crude. 1044 01:06:45,960 --> 01:06:48,280 It's not set in stone. 1045 01:06:48,280 --> 01:06:50,200 It's imperfect. 1046 01:06:50,200 --> 01:06:51,880 It's just a starting point. 1047 01:06:51,880 --> 01:06:54,200 So I'm just putting that out there. 1048 01:06:54,200 --> 01:06:57,000 Just this is what we have. 1049 01:06:57,000 --> 01:06:59,560 So I think it's better than nothing. 1050 01:06:59,560 --> 01:07:03,160 I think it can definitely get a whole lot better. 1051 01:07:03,160 --> 01:07:06,600 So definitely hedging it. 1052 01:07:06,600 --> 01:07:13,800 So now you saw the process and you can see, OK, 1053 01:07:13,800 --> 01:07:16,200 there's a lot of assumptions built into this. 1054 01:07:16,200 --> 01:07:18,360 The model is imperfect. 1055 01:07:18,360 --> 01:07:21,560 The data is imperfect. 1056 01:07:21,560 --> 01:07:24,400 But it's all transparent. 1057 01:07:24,400 --> 01:07:26,840 I want to say that there's a lot to get your teeth into with 1058 01:07:26,840 --> 01:07:27,840 this. 1059 01:07:27,840 --> 01:07:28,840 Yes. 1060 01:07:28,840 --> 01:07:32,680 And so instead of just seeing, oh, $31 billion 1061 01:07:32,680 --> 01:07:35,440 and just having to take that at the expense of the state, 1062 01:07:35,440 --> 01:07:38,080 just having to take that at face value, 1063 01:07:38,080 --> 01:07:42,800 you can dive into all the different layers that 1064 01:07:42,800 --> 01:07:43,880 are involved. 1065 01:07:43,880 --> 01:07:45,560 You've got to have a baseline somewhere. 1066 01:07:45,560 --> 01:07:48,160 So here we are. 1067 01:07:48,160 --> 01:07:49,120 Exactly. 1068 01:07:49,120 --> 01:07:51,960 And the cool thing about forecasting 1069 01:07:51,960 --> 01:07:56,440 is we can check it, if we can get the data, that is. 1070 01:07:56,440 --> 01:08:03,200 So as 2022 unfolds, we can see how right or wrong we were 1071 01:08:03,200 --> 01:08:05,400 and adjust our model. 1072 01:08:05,400 --> 01:08:07,160 And that's part of the process. 1073 01:08:07,160 --> 01:08:10,520 So we've got to learn from our mistakes. 1074 01:08:10,520 --> 01:08:13,600 And inevitably, we've made some mistakes. 1075 01:08:13,600 --> 01:08:17,560 So the model can be improved upon. 1076 01:08:17,560 --> 01:08:20,400 And we can keep at it. 1077 01:08:20,400 --> 01:08:25,120 So for example, as January comes, some of these states 1078 01:08:25,120 --> 01:08:27,360 are coming online. 1079 01:08:27,360 --> 01:08:39,280 What was the state that just permitted here? 1080 01:08:39,280 --> 01:08:40,680 Delaware. 1081 01:08:40,680 --> 01:08:43,080 Yeah, Delaware and, oh, yes, Montana 1082 01:08:43,080 --> 01:08:45,600 just permitted adult use. 1083 01:08:45,600 --> 01:08:49,440 So as we see these various states unfolding 1084 01:08:49,440 --> 01:08:52,120 and collect more data, then each month 1085 01:08:52,120 --> 01:08:56,000 we can revise our forecasts for each month ahead. 1086 01:08:56,000 --> 01:08:58,560 So we can measure ourselves as we go 1087 01:08:58,560 --> 01:09:02,480 and then make better and better forecasts for the future. 1088 01:09:02,480 --> 01:09:06,720 So it's the best we can do. 1089 01:09:06,720 --> 01:09:09,160 I'd just like to thank you all for coming. 1090 01:09:09,160 --> 01:09:11,800 Thank you all for staying a little extra. 1091 01:09:11,800 --> 01:09:13,240 I know your time's precious. 1092 01:09:13,240 --> 01:09:16,960 So I'll go ahead and wrap it up here. 1093 01:09:16,960 --> 01:09:18,440 Thank you all for coming, though. 1094 01:09:18,440 --> 01:09:18,960 Thank you. 1095 01:09:18,960 --> 01:09:19,760 See you. 1096 01:09:19,760 --> 01:09:20,360 It was great. 1097 01:09:20,360 --> 01:09:21,080 Thank you. 1098 01:09:21,080 --> 01:09:22,040 Awesome. 1099 01:09:22,040 --> 01:09:23,360 Have a productive week. 1100 01:09:23,360 --> 01:09:26,280 Thank you for notes to the grindstone, and I'll see you next week. 1101 01:09:26,280 --> 01:09:53,920 Thank you.