Mitsuo Shiota 2019-04-19
- Summary
- Get meta data
- Get data
- Confirm the Hamamatsu-city’s come-back win in dumplings
- Hamamatsu-city also regained championship in grilled eel
- Persistent victory items are mostly speciality goods
- Who are the top rank earners in 2023?
- Ranks distribution in each city in 2023
Updated: 2024-02-06
I have made shinyapps.io, based on these data.
Warning: If you can’t read Japanese, it would be very difficult for you to follow me in this project of exploring e-Stat, Statistics of Japan. I tried to refer to English pages, but often failed to find them. Even in API, when I set the parameter lang as “E” for English, the response was “statsDataId equal to [0003125169] does not exist”. I had to set lang as “J” for Japanese, instead.
I hear which city in Japan consumed most dumplings (gyouza in Japanese, and jiaozi in Chinese) per household last year. This is an annual routine news report. In 2023, Hamamatsu-city came back to the top rank (Japanese), based on Family Income and Expenditure Survey by the Statistics Bureau, Ministry of Internal Affairs and Communications. So I get data from e-Stat by using estatapi package by yutannihilation.
I find that persistent winning items are speciality goods of each city. I list the top earners in 2023. I draw ranks distribution in each city in 2023. And I argue ranks distribution is partly due to total expenditure differences.
I had to register and get appID, following the instructions in this page (Japanese). I saved appID as ESTAT_API_KEY in .Renviron file.
estat_api_key <- Sys.getenv("ESTAT_API_KEY")I search around in Family Income and Expenditure Survey, and manage to know the statsDataId of the appropriate table is “0003348239” from this page (Japanese). As I mentioned in the warning in the top, I failed to set lang = “E”, so I set “J” instead.
Look at the structure of the response. It is a list of 6 data frames. estat_getStatsData function, which I will use later, utilizes @code and @name in each data frame, and add @name. In this case, I would like to keep @level in cat01. I make lookup table for cat01, and drop “@” in the column names. I can check which level I should use, and search “ぎょうざ”.
meta_info <- estatapi::estat_getMetaInfo(appId = estat_api_key,
lang = "J",
statsDataId = "0003348239")
str(meta_info)## List of 6
## $ tab : tibble [1 × 3] (S3: tbl_df/tbl/data.frame)
## ..$ @code : chr "01"
## ..$ @name : chr "金額"
## ..$ @level: chr ""
## $ cat01 : tibble [690 × 5] (S3: tbl_df/tbl/data.frame)
## ..$ @code : chr [1:690] "000100000" "000200000" "000300000" "000400000" ...
## ..$ @name : chr [1:690] "世帯数分布(抽出率調整)" "集計世帯数" "世帯人員" "18歳未満人員" ...
## ..$ @level : chr [1:690] "1" "1" "1" "2" ...
## ..$ @unit : chr [1:690] "一万分比" "世帯" "人" "人" ...
## ..$ @parentCode: chr [1:690] NA NA NA "000300000" ...
## $ cat02 : tibble [4 × 3] (S3: tbl_df/tbl/data.frame)
## ..$ @code : chr [1:4] "03" "04" "01" "02"
## ..$ @name : chr [1:4] "二人以上の世帯（2000年～）" "二人以上の世帯のうち勤労者世帯（2000年～）" "二人以上の世帯（農林漁家世帯を除く）（1985年～2007年,2017年）" "二人以上の世帯のうち勤労者世帯（農林漁家世帯を除く）（1985年～2007年,2017年）"
## ..$ @level: chr [1:4] "1" "1" "1" "1"
## $ area : tibble [53 × 3] (S3: tbl_df/tbl/data.frame)
## ..$ @code : chr [1:53] "00000" "01003" "02003" "03003" ...
## ..$ @name : chr [1:53] "全国" "01100 札幌市" "02201 青森市" "03201 盛岡市" ...
## ..$ @level: chr [1:53] "1" "1" "1" "1" ...
## $ time : tibble [39 × 3] (S3: tbl_df/tbl/data.frame)
## ..$ @code : chr [1:39] "1985000000" "1986000000" "1987000000" "1988000000" ...
## ..$ @name : chr [1:39] "1985年" "1986年" "1987年" "1988年" ...
## ..$ @level: chr [1:39] "1" "1" "1" "1" ...
## $ .names: tibble [5 × 2] (S3: tbl_df/tbl/data.frame)
## ..$ id : chr [1:5] "tab" "cat01" "cat02" "area" ...
## ..$ name: chr [1:5] "表章項目" "品目分類（2020年改定）" "世帯区分（年次－二人以上の世帯）" "地域区分" ...
lookup_cat01 <- meta_info$cat01
names(lookup_cat01) <- str_sub(names(lookup_cat01), start = 2L)
lookup_cat01 <- lookup_cat01 %>%
rename(cat01_code = code)
lookup_cat01 %>%
filter(level == 3)## # A tibble: 96 × 5
## cat01_code name level unit parentCode
## <chr> <chr> <chr> <chr> <chr>
## 1 000600000 65歳以上無職者人員 3 人 000500000
## 2 010110001 1.1.1 米 3 円 010100000
## 3 010120000 1.1.2 パン 3 円 010100000
## 4 010130000 1.1.3 麺類 3 円 010100000
## 5 010140000 1.1.4 他の穀類 3 円 010100000
## 6 010210000 1.2.1 生鮮魚介 3 円 010200000
## 7 010220000 1.2.2 塩干魚介 3 円 010200000
## 8 010230000 1.2.3 魚肉練製品 3 円 010200000
## 9 010240000 1.2.4 他の魚介加工品 3 円 010200000
## 10 010310000 1.3.1 生鮮肉 3 円 010300000
## # ℹ 86 more rows
lookup_cat01 %>%
filter(str_detect(name, "ぎょうざ"))## # A tibble: 1 × 5
## cat01_code name level unit parentCode
## <chr> <chr> <chr> <chr> <chr>
## 1 010920070 371 ぎょうざ 5 円 010920000
As the maximum records per GET is set to 100000, I must repeat five times in this case if I GET manually. Fortunately estat_getStatsData function automates this process. I add “level” of cat01, and clean up.
# get 599881 records
content <- estatapi::estat_getStatsData(
appId = estat_api_key,
lang = "J",
statsDataId = "0003348239")## Fetching record 1-100000... (total: 674401 records)
## Fetching record 100001-200000... (total: 674401 records)
## Fetching record 200001-300000... (total: 674401 records)
## Fetching record 300001-400000... (total: 674401 records)
## Fetching record 400001-500000... (total: 674401 records)
## Fetching record 500001-600000... (total: 674401 records)
## Fetching record 600001-674401... (total: 674401 records)
# add level column
temp <- lookup_cat01 %>%
select(cat01_code, level)
expenditure <- content %>%
left_join(temp, by = "cat01_code") %>%
mutate(
year = str_sub(time_code, start = 1L, end = 4L) %>% as.numeric()
) %>%
rename(city = 地域区分)According to this page (Japanese), the sample numbers per city are approximately 100 except Tokyo (408 samples). If I assume the standard deviation is 30 percent of the mean, the standard error ratio (ratio of standard deviation of mean estimates to the estimated mean) is 3 percent. Table 1-1 in this page publish the standard error ratios in cat01 of level less than 3 in 2013 survey, and the numbers do not contradict my assumption of 3 percent standard error ratio for cat01 of level 5.
## # A tibble: 10 × 2
## city value
## <chr> <dbl>
## 1 22130 浜松市 4041
## 2 45201 宮崎市 3498
## 3 09201 宇都宮市 3199
## 4 26100 京都市 2661
## 5 25201 大津市 2489
## 6 29201 奈良市 2440
## 7 23100 名古屋市 2390
## 8 11100 さいたま市 2287
## 9 13100 東京都区部 2204
## 10 27100 大阪市 2169
Miyazaki-city won in 2021 and 2022. Hamamatsu-city won in 2023. The battle between Utsunomiya-city and Hamamatsu-city has a long history. Let us draw all available years of top 10 cities in 2023.
By the way, these expenditures don’t include dumplings you eat at the restaurants.
Hamamatsu-city had been the top of eating the grilled eel in 11 years in a row up to 2018. It lost the top position in 2019, regained it in 2020, lost again in 2021, remained second in 2022, and regained it in 2023.
I search eel to get cat01_code. It is “010920010”.
## # A tibble: 1 × 5
## cat01_code name level unit parentCode
## <chr> <chr> <chr> <chr> <chr>
## 1 010920010 364 うなぎのかば焼き 5 円 010920000
Next, I rank 52 cities every year in 488 items of level 5. 52 cities consist of 47 prefectural capital cities and 5 non-prefectural-capital large cities whose population is more than half million.
I show the top 3 in eel every year.
## [,1] [,2] [,3]
## ranks "1" "2" "3"
## 2007 "21201 岐阜市" "23100 名古屋市" "26100 京都市"
## 2008 "22130 浜松市" "32201 松江市" "29201 奈良市"
## 2009 "22130 浜松市" "26100 京都市" "22100 静岡市"
## 2010 "22130 浜松市" "26100 京都市" "24201 津市"
## 2011 "22130 浜松市" "26100 京都市" "29201 奈良市"
## 2012 "22130 浜松市" "30201 和歌山市" "27100 大阪市"
## 2013 "22130 浜松市" "26100 京都市" "14130 川崎市"
## 2014 "22130 浜松市" "26100 京都市" "23100 名古屋市"
## 2015 "22130 浜松市" "17201 金沢市" "13100 東京都区部"
## 2016 "22130 浜松市" "25201 大津市" "23100 名古屋市"
## 2017 "22130 浜松市" "26100 京都市" "25201 大津市"
## 2018 "22130 浜松市" "26100 京都市" "17201 金沢市"
## 2019 "25201 大津市" "26100 京都市" "22100 静岡市"
## 2020 "22130 浜松市" "25201 大津市" "24201 津市"
## 2021 "26100 京都市" "22130 浜松市" "13100 東京都区部"
## 2022 "26100 京都市" "22130 浜松市" "17201 金沢市"
## 2023 "22130 浜松市" "22100 静岡市" "26100 京都市"
I count the number of cities which get the top rank in 17 years from 2007 to 2023 for each item.
There are 19 items where one city keeps the top rank. Most of these items match speciality goods in each city.
## # A tibble: 19 × 3
## cat01_code `品目分類（2020年改定）` city
## <chr> <chr> <chr>
## 1 010130010 130 生うどん・そば 37201 高松市
## 2 010211010 170 まぐろ 22100 静岡市
## 3 010211040 174 かつお 39201 高知市
## 4 010212020 192 しじみ 32201 松江市
## 5 010212040 194 ほたて貝 02201 青森市
## 6 010230010 203 揚げかまぼこ 46201 鹿児島市
## 7 010230020 204 ちくわ 31201 鳥取市
## 8 010230030 205 かまぼこ 04100 仙台市
## 9 010240010 210 かつお節・削り節 47201 那覇市
## 10 010240040 216 魚介の缶詰 47201 那覇市
## 11 010320040 229 他の加工肉 47201 那覇市
## 12 010512010 250 さつまいも 36201 徳島市
## 13 010512050 254 にんじん 47201 那覇市
## 14 010513010 260 さやまめ 15100 新潟市
## 15 010530020 281 油揚げ・がんもどき 18201 福井市
## 16 010540010 290 こんにゃく 06201 山形市
## 17 010610060 305 梨 31201 鳥取市
## 18 010800040 343 カステラ 42201 長崎市
## 19 011211010 390 日本そば・うどん 37201 高松市
In how many items among 488 did each city get the top in 2023?
## # A tibble: 52 × 2
## # Groups: city [52]
## city n
## <chr> <int>
## 1 11100 さいたま市 24
## 2 13100 東京都区部 23
## 3 39201 高知市 18
## 4 06201 山形市 16
## 5 02201 青森市 15
## 6 07201 福島市 15
## 7 26100 京都市 15
## 8 05201 秋田市 14
## 9 20201 長野市 14
## 10 22130 浜松市 14
## 11 16201 富山市 13
## 12 23100 名古屋市 13
## 13 27140 堺市 13
## 14 03201 盛岡市 12
## 15 15100 新潟市 12
## 16 17201 金沢市 12
## 17 31201 鳥取市 12
## 18 14130 川崎市 11
## 19 14150 相模原市 11
## 20 08201 水戸市 10
## 21 14100 横浜市 10
## 22 25201 大津市 10
## 23 47201 那覇市 10
## 24 09201 宇都宮市 9
## 25 40100 北九州市 9
## 26 40130 福岡市 9
## 27 21201 岐阜市 8
## 28 24201 津市 8
## 29 33100 岡山市 8
## 30 34100 広島市 8
## 31 01100 札幌市 7
## 32 04100 仙台市 7
## 33 18201 福井市 7
## 34 42201 長崎市 7
## 35 43100 熊本市 7
## 36 10201 前橋市 6
## 37 19201 甲府市 6
## 38 22100 静岡市 6
## 39 30201 和歌山市 6
## 40 32201 松江市 6
## 41 36201 徳島市 6
## 42 37201 高松市 6
## 43 41201 佐賀市 6
## 44 44201 大分市 6
## 45 28100 神戸市 5
## 46 46201 鹿児島市 5
## 47 12100 千葉市 3
## 48 29201 奈良市 3
## 49 35203 山口市 3
## 50 45201 宮崎市 2
## 51 27100 大阪市 1
## 52 38201 松山市 1
I draw the histogram that shows each city’s ranks of 488 items in 2023. Naha-city and other Kyushu cities and Wakayama-city tend to rank low, while Tokyo and other Kanto cities tend to rank high.
This pattern
reflects that total expenditures are high in Kanto, and low in
Wakayama-city, Naha-city and other Kyushu cities, partly due to price
level differences.
## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
EOL