Market Capitalization in R

As promised, we are just introducing the package and brushing up our ggplot skills. Therefore, we are focusing in this post on the evolution of market cap, allowing us to compare trends between different stocks. It is well known that growth stocks have experienced a massive surge in recent years, ranging from technology growth companies to pump-and-dump penny stocks, ‘meme-stocks’ defying the establishment (good luck Credit Suisse, SVB..) and SPACs frenzy. The market has been flooded with free money for almost a decade, with tech players lowering the entry barriers for retail investors, and benefiting from lockdown savings and subsidies.

Let’s examine the impact of this phenomenon on a set of stocks categorized as ‘growth’ and ‘value’. Since life is a continuum, we will first categorize them by hand according to the consensus. We will then conduct an initial analysis, which will be elaborated further in subsequent posts, attempting to cluster them based on their actual performance. In this chapter, we will use the market cap fluctuation to achieve this, although other features could also be utilized and macroeconomic factors taken into account. Stay tuned for upcoming TypeThePipe posts!

symbols = c('AAPL', 'META', 'AMZN', 'GOOG','MSFT')
data <- fmpcloudr::fmpc_security_mrktcap(symbols, limit = 30*12*12)
data %<>% mutate(date = ymd(date))
ggplot(data, aes(date, marketCap, colour = symbol)) +
 geom_line() 

Financial Times Theme in Ggplot

Not bad for a 2 lines plot, but lets add some Financial Times theme kind of flavor:

# old chunk in my codebase seen in Reddit but and can't find the source. If anyone has the original link, very glad to add it here! :)
ft_theme <- 
 theme(panel.grid.major.x =element_line(colour = "wheat4"),
 panel.grid.major.y =element_line(colour = "wheat4"),
 panel.grid.minor.x = element_blank(),
 panel.grid.minor.y = element_blank(),
 axis.text = element_text(colour = "black", size = 8, face = "bold"),
 axis.title.x = element_text(colour = "black", size = 8, face = "bold", vjust = 0.8),
 axis.title.y = element_blank(),
 strip.text = element_text(size=8, colour = "blue4", face = "bold"),
 plot.title = ggtext::element_markdown(color = "black", size = 14, face = "bold"),
 plot.subtitle = ggtext::element_markdown(color = "black", size = 8),
 plot.tag = element_text(color = "black", face = "italic", size = 182, lineheight = 0.9),
 plot.tag.position = c(0.15,0.02),
 plot.caption = ggtext::element_markdown(),
 panel.background = element_rect(fill = "seashell2"),
 plot.background = element_rect(fill = "seashell2"), 
 panel.border = element_blank(),
 panel.spacing.y = unit(3, "lines")
 ) 
data %>% 
 ggplot(aes(date, marketCap, group=symbol, colour=symbol)) +
 geom_line(size = 0.5) + 
 guides(colour="none") + 
 geom_label(data = data %>% filter(date==max(date)),
 aes(label = symbol),
 label.size = 0.1, size=2.2,
 label.padding = unit(0.2, "lines"),
 nudge_x=200) +
 coord_cartesian(clip = 'off') + # Allow labels to bleed past the canvas boundaries
 scale_x_date(date_breaks = "6 month", # Date labels for each month.
 date_labels = "%b%y",
 guide = guide_axis(angle = 90)) +
 scale_y_continuous(labels=function(x) sprintf("%d Billion", x / 10^9)) + 
 labs(title="Market Cap evolution - FAAMG",
 subtitle="<span style='color: #A9A9A9;'><b>Apple</b></span>,
 <span style='color: #E7B800;'><b>Amazon</b></span>, 
 <span style='color: #FC4E07;'><b>Microsoft</b></span> and 
 <span style='color: #0D7680;'><b>Meta</b></span>",
 x=" ")+ 
 scale_color_manual(values = c("#A9A9A9", "#E7B800", "#34AB53", "#0D7680", "#FC4E07")) +
 ft_theme

Let’s do the same reusing our Financial Times inspired theme to some considered value stocks.

symbols_value = c('MCD', 'BRK-A', 'PG', 'KO')
data_value <- fmpcloudr::fmpc_security_mrktcap(symbols_value, limit = 30*12*12)
data_value %<>% mutate(date = ymd(date),
 symbol = str_replace(symbol, "BRK-A", "BRK"))
data_value %>% 
 ggplot(aes(date, marketCap, group=symbol, colour=symbol)) +
 geom_line(size = 0.5) + 
 guides(colour="none") + 
 geom_label(data = data_value %>% filter(date==max(date)),
 aes(label = symbol),
 label.size = 0.1, size=3,
 label.padding = unit(0.2, "lines"),
 nudge_x=200) +
 coord_cartesian(clip = 'off') + # Allow labels to bleed past the canvas boundaries
 scale_x_date(date_breaks = "6 month", # Date labels for each month.
 date_labels = "%b%y",
 guide = guide_axis(angle = 90)) +
 scale_y_continuous(labels=function(x) sprintf("%d Billion", x / 10^9)) + 
 labs(title="Market Cap evolution - Value Stocks",
 subtitle="<span style='color: #554178;'><b>Berkshire Hathaway</b></span>,
 <span style='color: #003CAE;'><b>Procter & Gamble</b></span>, 
 <span style='color: #df2b24;'><b>Coca-Cola</b></span> and 
 <span style='color: #E7B800;'><b>Mcdonald's</b></span>",
 x=" ")+ 
 scale_color_manual(values = c("#554178", "#df2b24", "#E7B800", "#003CAE")) +
 ft_theme

Could we cluster stocks based in their recent 2023 performance?

Time series clustering is a very controversial topic. Therefore we are skipping it during this first introductory post, and we are using a simple clustering based on simple stats: daily return (approximated using the already downloaded market cap delta) and volatility.

data_basic_stats <- data %>% 
 filter(date>='2023-01-01') %>% 
 group_by(symbol) %>% 
 mutate(lag_mktcap = lag(marketCap,order_by = date),
 diff = (marketCap - lag_mktcap) /lag_mktcap * 100) %>% 
 summarise_at(.vars = 'diff', c(funs(mean(., na.rm = TRUE)), funs(sd(., na.rm = TRUE))))
data_value_basic_stats <- data_value %>% 
 filter(date>='2023-01-01') %>% 
 group_by(symbol) %>% 
 mutate(lag_mktcap = lag(marketCap,order_by = date),
 diff = (marketCap - lag_mktcap) /lag_mktcap * 100) %>% 
 summarise_at(.vars = 'diff', c(funs(mean(., na.rm = TRUE)), funs(sd(., na.rm = TRUE))))

Since we are only examining the dataset and it only has two dimensions, we opt to use the straightforward elbow method to select 3 clusters. However, we could also try with 2 and 4 clusters to compare the outcomes.

data_stats <- data_basic_stats %>% bind_rows(data_value_basic_stats)

tot_withinss <- map_dbl(2:5, function(k) {
 model <- kmeans(data_stats %>% select(mean, sd), centers = k)
 model$tot.withinss
})

data.frame(k = 2:5,
 tot_withinss = tot_withinss) %>%
 ggplot(aes(x = k, y = tot_withinss)) +
 geom_line() +
 scale_x_continuous(breaks = 2:5)

fit = kmeans(data_stats %>% select(mean, sd), 3)

data_stats %<>% mutate(cluster = fit$cluster,
 cluster_mean = fit$centers[cluster,1],
 cluster_sd = fit$centers[cluster,2],
 )

In the field of economics, there is a tendency to apply discrete categories, often based on historical but arbitrary thresholds, such as recession / not recession, stagflation, growth stocks versus value stocks… We should go beyond basic KPIs and try to approximate underlaying factors driving the markets.

Throughout this series, we will strive to analyze and categorize data in a way that reflects richer and higher-dimensional information.

ggplot(data_stats, aes(mean, sd, colour=factor(cluster))) +
 geom_point() +
 geom_label(aes(label = symbol), size=3, label.r = unit(5, 'points')) + 
 guides(colour=FALSE)

We can easily expand our analysis to other stocks. Let’s discover how to do so in using the Financial Market Prep API and R package, and in next post we’ll elaborate on those ts clusterization methods, common patterns, turning points…

total_tickers <- fmpc_symbols_available()
total_tickers$exchangeShortName %>% unique() %>% head(20)
## [1] "MCX" "NSE" "STO" "HEL" "CPH" "NASDAQ"
## [7] "TAL" "CBT" "EURONEXT" "LSE" "IST" "VIE"
## [13] "JNB" "MIL" "XETRA" "AMS" "PRA" "BSE"
## [19] "LIS" "OSE"
list_symbols <- total_tickers %>% filter(exchangeShortName=='NYSE') %>% pull(symbol)

This APi call provides quite a lot (all) of them!

data <- fmpcloudr::fmpc_security_mrktcap(list_symbols, limit = 30*3)

We now have the capability to conduct the analysis on the entire NYSE, or at least the most highly capitalized stocks. But that will be another day 😉

For now, we can see again a decay in the value stocks in these two first months of 2023 vs the growth ones. The market has once again received a dose of optimism. Could it be a good option to start/continue a DCA strategy? Perhaps its suitability varies based on industry and current valuation? Most likely, it does.

We are extending this analysis and talking about valuation metrics in our second post of the series! So…