An Introduction to Analyzing the NAWS Public Access Data

An Introduction to Analyzing the

NAWS Public Access Data

July 2020

Release 1.1

 
Contents 
Overview on Conducting Analysis with the NAWSPAD  .................................................. 3 
Software Options for Analyzing the NAWS Data  ......................................................... 3 
Calculate Weighted Means and Proportions  ............................................................... 4 
Calculate Design-Corrected Standard Errors  .............................................................. 4 
Additional Documentation to Support NAWSPAD Users  ............................................ 4 
How to Calculate Weighted Means and Weighted Proportions using Excel  ................... 6 
Download the NAWS Excel files  ................................................................................. 6 
Apply Sampling Weights  ............................................................................................. 6 
Calculate Weighted Means and Proportions  ............................................................... 6 
Calculate Weighted Means  ......................................................................................... 7 
Calculate Weighted Means for All Fiscal Years  ....................................................... 7 
Calculate Weighted Means for Specific Fiscal Years  .............................................. 8 
Calculate Weighted Proportions ................................................................................ 10 
Calculate Weighted Proportions for All Fiscal Years  ............................................. 10 
Calculate Weighted Proportions for Specific Fiscal Years  ..................................... 16 

Overview on Conducting Analysis with the NAWSPAD

This document provides an introduction to the NAWS Public Access Data (NAWSPAD),

alerting the user to statistical issues related to the National Agricultural Workers

Survey’s (NAWS) complex sampling design and outlining available options for

addressing them. The NAWS uses a stratified multi-stage sampling design to account

for seasonal and regional fluctuations in the level of employment in crop agriculture. The

NAWS has seven levels of sampling. The first two levels are to create 36 strata

determined by three interviewing cycles and 12 agricultural regions. The next level

consists of the primary sampling unit (PSU) within each stratum. The PSU is the county

cluster (Farm Labor Area). The next four levels of sampling occur within the county

cluster. Within each county cluster, the NAWS draws random samples of counties, ZIP

Code regions, and employers. At the final level of sampling, workers are randomly

sampled within employers. Further information on the NAWS sampling design is

available in the Statistical Methods of the National Agricultural Workers Survey available

on the NAWS website (http://www.doleta.gov/agworker/naws.cfm).

The NAWS’s multi-level sampling design adds complexity to data analysis in two ways:

1. To produce accurate statistics, users need to apply sampling and post-sampling

weights to account for differences in sampling probabilities and to correct for non-

response at the region and cycle levels.

2. Users who wish to calculate the standard error of point estimates need to

account for the complex multi-stage sampling design. The NAWSPAD supports

using the balanced repeated replication (BRR) method. This method does not

require information on the strata or PSU, which are suppressed from the

NAWSPAD for privacy reasons.

Software Options for Analyzing the NAWS Data

The NAWSPAD files are available in SAS, Microsoft Excel, and CSV (comma separated

values) formats. These files can be read into other data analysis programs, such as

SPSS and Stata. NAWS users who do not have access to analysis software such as

SAS, SPSS, R, or Stata, or users who only need to quickly calculate the mean or

proportion of specific variables, can perform these calculations using the NAWS Excel

files. In addition, there are several documents to support users, including codebooks,

questionnaires, and explanation of the NAWS methodology. A list of these documents

is provided below.

Calculate Weighted Means and Proportions

Users with access to statistical software such as SAS, SPSS, R, or Stata should

consult the respective manuals for how to calculate weighted means and proportions

using the weighting variable PWTYCRD. MS Excel users can calculate weighted

means and proportions using the instructions in Chapter 1 “How to Calculate

Weighted Means and Weighted Proportions from NAWS Public Access Data Using

Excel.” This chapter provides step-by-step instructions on performing the

calculations. The process can be tedious if analyzing many variables but does

provide the same means and proportions produced by statistical software.

Calculate Design-Corrected Standard Errors

This document provides specific guidance and detailed explanations on how to

calculate weighted means and weighted proportions using excel.

Additional Documentation to Support NAWSPAD Users

The following additional documentation about the NAWS and the NAWSPAD

may be useful. These documents are available on the NAWS website

(http://www.doleta.gov/agworker/naws.cfm).

•

NAWSPAD Variables and Labels

•

The NAWSPAD codebook

•

Supporting statement [Part A of the Paperwork Reduction Act (PRA)

Information Collection Request (ICR)]

•

Statistical Methods of the National Agricultural Workers Survey (Part B of

the PRA ICR)

•

Field sampling protocols

•

Questionnaires in English and Spanish

•

For users interested in regional analysis:

o Sampling regions

o Correspondence between NAWS and USDA Farm Labor Survey

sampling regions

o Analysis regions in the NAWSPAD file

How to Calculate Weighted Means and Weighted

Proportions using Excel

This chapter provides step-by-step instruction on calculating weighted means and

proportions using Excel.

Step 1

Download the NAWS Excel files

From the “Public Data Files” link on the NAWS website

(http://www.doleta.gov/agworker/naws.cfm), download the two files that are available in

Excel format: NAWS_A2E185 (which contains the first half of the variables) and

NAWS_F2Y185 (which contains the second half of the variables).

Step 2

Apply Sampling Weights

The NAWS uses a complex sampling design that includes both stratification and

clustering; for this reason, users must make use of sampling weights to adjust the

relative value of each farmworker so that population estimates may be obtained from

the sample. The NAWS sampling weight variable PWTYCRD includes a factor which

correctly proportions the data for analysis. The PWTYCRD variable is used for almost all

NAWS analysis and allows merging several years of data together. At least two

consecutive years of data should be combined to obtain robust. The PWTYCRD

variable can be found in both the NAWS_A2E185 and NAWS_F2Y185 files. Further

information on how the NAWS sampling weight is calculated is provided in Statistical

Methods of the National Agricultural Workers Survey, available on the NAWS website

(http://www.doleta.gov/agworker/naws.cfm).

Step 3

Calculate Weighted Means and Proportions

There are two types of variables in the NAWSPAD: continuous and categorical

variables. Users can calculate weighted means for continuous variables using the

instructions in Step 3a, and weighted proportions for categorical variables using the

instructions in Step 3b. Examples of continuous variables include age of farmworker,

wages, number of work days, and numbers of years in the United States. Categorical

variables include those for which each farmworker respondent is assigned a value

indicating his/her membership in one of several possible categories. Examples of

categorical variables include gender, employment status, and legal status. Users should

consult the codebook for more information on the variables contained in the NAWSPAD.

Step 3a

Calculate Weighted Means

The mean, or average, of a variable is the sum of all values across all the observations

divided by the number of observations. This sample mean assumes that each

observation in the sample has an equal weight. However, when calculating the mean of

a variable in the NAWSPAD, users need to apply the sampling weight variable

PWTYCRD to adjust for the relative value of each farmworker in the sample, because

each farmworker contributes differently to the final mean depending on his/her sampling

weight.

The weighted mean can be calculated in Excel with a formula that uses both the

SUMPRODUCT and SUM functions. For example, if users want to calculate the mean

age of farmworkers then the SUMPRODUCT function is used to calculate the numerator

in the formula, which is the sum of the products of each farmworker’s age and his/her

sampling weight (PWTYCRD). The SUM function is used to calculate the denominator in

the formula, which is the sum of all farmworker sampling weights (PWTYCRD).

Calculate Weighted Means for All Fiscal Years

The following instructions show how to calculate the weighted mean of a continuous

variable for all fiscal years in the dataset combined in Excel. For the purpose of

illustration, we will use AGE as the variable of interest and we assume that the Excel

data set contains 500 farm workers interviewed in all fiscal years.

1. Determine the variable(s) for which you would like to calculate the weighted

mean(s). In our example, we want to calculate mean AGE.

2. Locate the variable of interest and the PWTYCRD variable in the NAWS Excel

file. For our example, the variable AGE is located in column A and the variable

PWTYCRD is located in column B (see Figure 1). There are 500 rows in our

example dataset, one for each farmworker interviewed. The first row contains

the variable names.

3. The Excel formula for calculating a weighted mean is

SUMPRODUCT(array1,array2,array3,...)/SUM(number1, [number2], [number3],

...). Using our example, the formula to calculate farmworkers’ mean age is

SUMPRODUCT(A2:A501, B2:B501)/SUM(B2:B501).

a. SUMPRODUCT(A2:A501, B2:B501) calculates the numerator, which

sums the products of each farmworker’s age (AGE) and his/her sampling

weight (PWTYCRD).

b. SUM(B2:B501) calculates the denominator, which is the sum of all 500

farmworker sampling weights.

c. The full formula SUMPRODUCT(A2:A501, B2:B501)/SUM(B2:B501)

calculates the weighted mean age for the 500 farmworkers in the data set.

Figure 1 shows an illustration of the Excel data set with the formula written

in it.

Figure 1: Formula for Weighted Mean

Calculate Weighted Means for Specific Fiscal Years

The previous step described how to calculate weighted means for a continuous variable

using all fiscal years in the dataset. It is also possible to calculate weighted means for

specific years using the Fiscal Year (FY) variable. To ensure that your results are as

accurate as possible, you need to combine at least two consecutive fiscal years of data

(e.g., 2015 and 2016). For more information about combining fiscal years, please

consult the Statistical Methods of the National Agricultural Workers Survey, available on

the NAWS website (http://www.doleta.gov/agworker/naws.cfm), which describes the

statistical methods of the NAWS.

There are two options to create a worksheet containing data for specific fiscal years.

Option 1: Use the Filter option in Excel Before entering any formulas into your

worksheet, select all cells in the worksheet, go to the “Data” tab on the menu bar at the

top of the worksheet (see Figure 2), and click the “Filter” button. A small drop-down

arrow will appear next to each variable name. Click on the drop-down arrow next to the

“FY” variable name (see Figure 2). This will open a dialogue box and at the bottom you

will see a list of all the fiscal years available for selection (see Figure 2). Select the fiscal

years you are interested in and click “OK”. You can pick as many fiscal years as you

want, but you should pick at least two consecutive years. Now that you have applied the

filter, you can calculate weighted means using the formula described in the “Calculate

Weighted Means for All Fiscal Years” section above, steps 1 through 3.

Figure 2: Apply a Filter to "FY"

Option 2: Copy and Paste Before you do any calculations, copy the column headings

for all the variables you need (i.e., the variable of interest, PWTYCRD, and FY) and the

rows corresponding to the fiscal years you are interested in. Paste them into a new

worksheet. Make sure you copy all the rows included in the fiscal years you have

chosen; missing even one row can change your results. In the new worksheet, calculate

weighted means using the formula described in the “Calculate Weighted Means for All

Fiscal Years” section above, steps 1 through 3.

For more help with calculating weighted means in Excel, please consult the “How to

calculate weighted averages in Excel” page on the Microsoft website

(http://support.microsoft.com/kb/214049).

Step 3b

Calculate Weighted Proportions

When calculating weighted proportions for categorical variables, users must apply the

sampling weight variable PWTYCRD to adjust for the relative value of each farmworker

in the sample (see Step 2 Apply Sampling Weights at the beginning of this document).

The most efficient way to calculate weighted proportions in Excel is to first organize the

data using PivotTables. The following instructions use PivotTables to help you easily

calculate weighted proportions for all fiscal years combined or for specific fiscal years.

Calculate Weighted Proportions for All Fiscal Years

The instructions below describe the method for calculating weighted proportions for all

fiscal years in the dataset. For the purpose of illustration, we will use the variable

CURRSTAT (the farmworker’s legal status at the time of interview), which has four

categories coded with the numeric values 1 through 4. The calculation described here

will yield the weighted proportion for each of the four categories of CURRSTAT.

To calculate weighted proportions for all fiscal years combined:

1. Determine the variable(s) for which you would like to calculate the weighted

proportions. In our example, we want to calculate the weighted proportions of

CURRSTAT.

2. Highlight all the cells in your spreadsheet. Go to the “Insert” tab on the menu bar

at the top of the worksheet and select “PivotTable” (see Figure 3). The “Create

PivotTable” dialogue box will appear, and two options will automatically be

selected: “Select a table or range” (which shows the range of all the cells

selected in our spreadsheet) and “New Worksheet” (see Figure 3). Click “OK”.

Figure 3: Create a PivotTable

3. After clicking “OK”, a new worksheet will open. On the left side of the new

worksheet you will see an empty PivotTable. On the right side you will see a

PivotTable Field List from which you can select the variables that you want to

add to your PivotTable (see Figure 4). Using our example of CURRSTAT as the

variable of interest:

a. Drag CURRSTAT and drop it into the “Row Labels” area.

b. Drag PWTYCRD and drop it into the “Values” area; you will see “Count of

PWTYCRD” with a drop-down arrow next to it (see Figure 4).

Figure 4: PivotTable Field List

c. Click on the drop-down arrow next to “Count of PWTYCRD” and click

“Value Field Settings”. In the Value Field Settings dialogue box, select the

“Sum” option under “Summarize value field by” (see Figure 5).

Figure 5: Value Field Settings for PWTYCRD

d. Click “OK” to create your PivotTable. The left column in the table contains

the four categories of CURRSTAT. The right column contains the sum of

the PWTYCRD values for each CURRSTAT category (see Figure 6); this

is the column that you will use to calculate your weighted proportions.

Figure 6: Created PivotTable

Note: The columns in your PivotTable may have headings different from those pictured based on the

version of Excel you are using. The values in the right column will likely differ from those pictured,

depending on the fiscal years of data you are analyzing.

4. Now you are ready to calculate the weighted proportions of the variable (i.e., how

often each category of the variable occurs). For each category of the variable

(i.e., row in the PivotTable):

a. Create a formula that divides the value in the right column of the

PivotTable by the “Grand Total” (see Figure 7).

b. Convert the resulting values from decimals to percentages (see Figure 7).

Note: Because you are calculating weighted proportions based on cells

that were created using a PivotTable, you cannot use the fill handle to

apply the formula for the first category of the variable to the other

categories of the variable. If you want to be able to use the fill handle

option, copy the PivotTable and paste it elsewhere in your worksheet

using “Paste Special” with the “Values” option. This will remove all

embedded formulas and provide you with a static table that you can use

for your calculations.

Figure 7: Formula for Weighted Proportion

Note that because your formula for calculating weighted proportion uses cells in a

PivotTable, you cannot use the fill handle to apply the formula for the first category of

the variable to the other categories of the variable. If you want to be able to use the fill

handle option, copy the PivotTable and paste it elsewhere in your worksheet using

“Paste Special” with the “Values” option. This will remove all embedded formulas and

provide you with a static table that you can use for your calculations.

The category labeled “(blank)” represents the respondents who are missing a value for

the variable. If left in the PivotTable, the category of missing values will comprise a

share of respondents. If you want to exclude missing values from your analysis, you can

filter them out by clicking the down arrow next to CURRSTAT, unchecking the box next

to “(blank)” in the filter menu, then clicking “OK” (see Figure 8).

Figure 8: Filter “(blank)” from Created PivotTable

For additional help creating PivotTables in Excel, please consult the “Create or delete a

PivotTable or PivotChart report” page on the Microsoft website

(https://support.office.com/en-us/article/Create-or-delete-a-PivotTable-or-PivotChart-

report-d09e4d07-8cd6-4b60-afad-8fb67418800f?CorrelationId=213dd195-0deb-4f22-

8ec8-e80eca1c2fd0&ui=en-US&rs=en-US&ad=US).

Calculate Weighted Proportions for Specific Fiscal Years

The process for calculating weighted proportions for specific fiscal years is very similar

to the process for calculating weighted proportions for all fiscal years. To ensure that the

data results are as accurate as possible, you need to combine at least two consecutive

fiscal years of data (e.g., 2015 and 2016). For more information about combining fiscal

years, please consult the Statistical Methods of the National Agricultural Workers

Survey which describes the statistical methods of the NAWS.

There are two options for calculating weighted proportions for specific fiscal years.

Option 1: Use the Report Filter option in a PivotTable Create a PivotTable by

following steps 1 through 3 in the “Calculate Weighted Proportions for All Fiscal Years”

section above. After you have changed the “Value Field Settings” for PWTYCRD to

“Sum of PWTYCRD” as described in step 3, return to the PivotTable Field List on the

right side of the screen, locate the variable “FY”, and drag and drop it into the “Report

Filter” area below the list. Once you have done this, the variable “FY” will appear in the

“Report Filter” section above your created PivotTable (see Figure 9).

Figure 9: Apply “FY” to the Report Filter

The next step is to select the fiscal years you are interested in. Click the down arrow

next to (All) and check the “Select Multiple Items” box. A check box will appear to the

left of each Fiscal Year in the list; click in the check box to the left of (All) to deselect all

of them. Locate the years that you are interested in, check the box for each (we use

2001-2002, for the purpose of illustration), then click “OK” (see Figure 10).

Figure 10: Select Specific Fiscal Years

Once the report filter is applied, the sum of the PWTYCRD values for each category of

CURRSTAT and the Grand Total for the table will recalculate to reflect the weighted

frequencies for only those fiscal years which you have selected. The final step is to

calculate the weighted proportions according to the procedure described in step 4 of the

“Calculate Weighted Proportions for All Fiscal Years” section above.

Option 2: Copy and Paste Before you do any of the steps involved in creating a

PivotTable, copy the column headings for all the variables you need (i.e., the variable of

interest, PWTYCRD, and FY) and the rows corresponding to the fiscal years you are

interested in. Paste them into a new worksheet. Make sure you copy all the rows

included in the fiscal years you have chosen; missing even one row can change your

results. In the new worksheet, create a PivotTable and calculate weighted proportions

according to the process described in steps 1 through 4 of the “Calculate Weighted

Proportions for All Fiscal Years” section above.

For additional help creating PivotTables in Excel, please consult the “Create or delete a

PivotTable or PivotChart report” page on the Microsoft website

(https://support.office.com/en-us/article/Create-or-delete-a-PivotTable-or-PivotChart-

report-d09e4d07-8cd6-4b60-afad-8fb67418800f?CorrelationId=213dd195-0deb-4f22-

8ec8-e80eca1c2fd0&ui=en-US&rs=en-US&ad=US)